SOCIAL PROTECTION & JOBS DISCUSSION PAPER No. 2505 | MARCH 2025 A slippery slope: the opportunities and risks of digital approaches and technology in Social Protection Systems Phillippe Leite Andres Chamba Youssef Zaarour Dominique Leska-See1 1 This Working Paper was prepared by and inter-agency team – World Bank, World Food Program and Deutsche Gesellschaft für Internationale Zusammenarbeit (GIZ) – under the Social Protection Interagency Cooperation Board (SPIAC-B) guidance. Contacts: Phillippe Leite (pleite@worldbank.org), Andres Chamba (andres.chamba@wfp.org), Youssef Zaarour (youssef. zaarour@wfp.org) and Dominique Leska-See (dominique.leska-see@giz.de). © 2025 International Bank for Reconstruction and Development / The World Bank 1818 H Street NW, Washington DC 20433 Telephone: +1 (202) 473 1000; Internet: www.worldbank.org. This work is a product of the staff of The World Bank with external contributions. The findings, interpretations, and conclusions expressed in this work do not necessarily reflect the views of The World Bank, its Board of Executive Directors, or the governments they represent. The World Bank does not guarantee the accuracy of the data included in this work. The boundaries, colors, denominations, and other information shown on any map in this work do not imply any judgment on the part of The World Bank concerning the legal status of any territory or the endorsement or acceptance of such boundaries. RIGHTS AND PERMISSIONS The material in this work is subject to copyright. Because The World Bank encourages dissemination of its knowledge, this work may be reproduced, in whole or in part, for noncommercial purposes as long as full attribution to this work is given. Any queries on rights and licenses, including subsidiary rights, should be addressed to World Bank Publications, The World Bank Group, 1818 H Street NW, Washington, DC 20433, USA; fax: +1 (202) 522 2625; e-mail: pubrights@worldbank.org. A slippery slope: the opportunities and risks of digital approaches and technology in Social Protection Systems Phillippe Leite, Andres Chamba, Youssef Zaarour, and Dominique Leska-See 1 January 2025 Abstract Advances in technology have the potential to enhance social protection services delivery but come with risks, such as data privacy concerns, exclusion, and biases. To use digital technologies effectively, strong frameworks, infrastructure, and capacity are essential. Without these, technology may inadvertently harm rather than benefit the intended populations. Technology can improve access, outreach, training, monitoring, and secure payments, among others, but risks must be managed by clearly defining roles and responsibilities among stakeholders. Hence, policymakers must understand challenges and define processes before selecting technology, viewing it as a tool to complement, not replace, non-digital services. Therefore, technology should support human resources in social protection programs, and aligning innovation with effective safeguards can maximize its potential for equitable and responsible outcomes. JEL Classification: D60, D70, D80, D81, D83, I38, O33, O35, O38, and O57 Keywords: Artificial Intelligence, Big Data, machine learning, social registries (aka unified registries, single registries, unique registries, registration and eligibility systems), social protection delivery system, social protection system, delivery systems, delivery chain, social assistance, services, transfers, social protection, data security, data privacy, data collection. 1 This Working Paper was prepared by and inter-agency team – World Bank, World Food Program and Deutsche Gesellschaft für Internationale Zusammenarbeit (GIZ) – under the Social Protection Interagency Cooperation Board (SPIAC-B) guidance. Contacts: Phillippe Leite (pleite@worldbank.org), Andres Chamba (andres.chamba@wfp.org), Youssef Zaarour (youssef.zaarour@wfp.org) and Dominique Leska-See (dominique.leska-see@giz.de). 1 Contents I. Introduction..........................................................................................................3 II. The opportunities and gains from technology in the System .......................................5 III. The risks and challenges of technology in the System ................................................9 A. An overview ......................................................................................................9 B. A brief technical dive into risks ..........................................................................12 IV. Addressing these risks ..........................................................................................14 A. Institutional Arrangements................................................................................14 B. Political Environment .......................................................................................17 C. Legal dimension ...............................................................................................19 D. Data systems ...................................................................................................23 E. Minimal data collection and access ....................................................................26 F. Consent for disclosure and sharing of information ...............................................27 G. AI and Big Data crescent ...................................................................................28 H. Digital Exclusion ...............................................................................................33 I. Systemic Exclusion ...........................................................................................34 J. Digital literacy/exclusion (access) .......................................................................35 V. The importance of Human resources .....................................................................37 VI. Conclusion ..........................................................................................................39 Bibliography ...............................................................................................................41 2 I. Introduction 1. In today’s world, Social Protection Systems aim to promote social inclusion and provide access to various benefits and services. New and fast-growing technology plays a very important role in such systems. The potential of using technology as a catalyst for change in the sphere of social protection systems delves into various aspects, ranging from streamlining delivery, programming, operations, processes, and procedures to supporting tools and personnel that constitute Social Protection Systems. However, technology can be misused and brings risks to it. 2. A Social Protection System (henceforth System) offers a myriad of activities, including provision of contributory and non-contributory benefits, and of services, to meet the diverse needs of the populations. It relies on a set of interactions between the population in need and the institutions along the delivery chain. 2 3. This System embodies our overall objective, our raison d'être, and is formed by a comprehensive set of social benefits and services (e.g., (un)conditional cash transfers, unemployment assistance and benefits, disability assistance and benefits, social insurance, active labor market programs, childcare and old age care services, social pensions, unemployment insurance 3, labor market activation programs, old-age pensions, etc.). These activities operate in an organized manner, forming a cohesive network to serve a common purpose: producing the outputs and outcomes necessary to reach the desired social protection goals. 4. To operate in a cost-effective and transparent way, the System can greatly benefit from the integration of technology to establish an information system that transforms data into standardized formats or aggregations. This supports implementation, management, monitoring, and evaluation process. Therefore, technology serves as a crucial enabler, not the end goal, and acting as a tool that bridges the last mile gap between the government and the population for accessing available benefits and services. 2 Delivery Chain encompasses multiple phases during implementation of a social program including: (a) determining potential eligibility, via outreach; intake & registration; assessment of needs and conditions; (b) taking decisions on enrollment and the benefits or service package; and (c) carrying out the implementation cycle of transactions (payments or service provision) and active case management (including counseling, conditionalities monitoring, accompanying measures, grievance redress, and so forth). See more in Leite at al. (2017) and Lindert and al (2020). 3 See WFP (2021) and World Bank (2022). 3 5. More importantly, while technology enhances efficiency and transparency, it is essential to acknowledge and address the risks associated with the digitization process that supports the System. By being mindful of these risks, we can ensure that the integration of technology strengthens the System without compromising its integrity or accessibility. 6. This System can be broad enough to encompass both contributory and non- contributory schemes, as well as labor market activation programs. Alternatively, it can be broken down into multiple sub-systems that support user programs with shared specificities. These sub-systems are designed with common features that allow for their integration and interoperability. The System is then uniquely positioned to help individuals at any moment they need support. More specifically for the poor and vulnerable populations, the System helps in preparing for, coping with, and adapting to shocks that can negatively impact human capital investment in both children and livelihoods. 7. The System also helps governments to implement macroeconomic or sectoral reforms that support efficiency and pro-poor growth through for example revision of certain subsidies that are regressive. Interventions offered in such a System can enhance employability, skills and productivity when focusing to the youth and adult workforce, as well as facilitating access to better jobs, while protecting people against old-age poverty. The System also plays a multifaceted role, contributing not only towards climate adaptation but also to mitigation efforts against natural shocks, buffering against economic shocks and bolstering household resilience.4 8. To achieve its objectives, the System must be comprehensive, adequately funded, and cost-efficient to ensure access across various stages of the life cycle. It is imperative to reach individuals in an effective and transparent manner to enhance access to public services, while maintaining administrative efficiency at all levels to deliver quality public services equitably. 9. Leveraging digital approaches and technology, henceforth technology, as pivotal tools that link individuals to institutions, thereby facilitating the administration and delivery of services improves outcomes. By streamlining the value chain processes, digitization and technology can help simplify access, promoting the efficiency and effectiveness of service delivery. Therefore, the integration of digital solutions is a fundamental enabler to provide equitable, efficient, and transparent access to social protection benefits and services. 4 See Bowen et al. (2020) and World Bank (2023) 4 10. However, technology is not a panacea and not without risks. It complements, rather than replaces, non-digital services and support human resources. By aligning innovation with effective safeguards and investing in human resources, we can maximize the potential of Social Protection Systems to achieve equitable and responsible outcomes. This strategic approach ensures that while we harness the benefits of technology, we also mitigate its risks and reinforce the essential human elements of service delivery. II. The opportunities and gains from technology in the System 11. The role of technology is constantly evolving, with countries and organizations at different stages of technological integration in their social protection initiatives. Various factors contribute to this diversity, including resources (financial, human and others), skills levels, resistance to change (including fear of it in some cases), behavioral traits, knowledge, complexity, perceived value of technology, digitization efforts, infrastructure, digital assets (both physical and virtual or software-related), data, and awareness of processes and procedures and their linkages to and alignment with policy-programming, operations, and tools. 12. However, the rapid growth of technologies supporting the delivery of social programs is driven by the push to innovate and to transition away from individual and isolated interventions to the delivery of broader, harmonized and coordinated social benefits and services to the population under a System. 5 Increasingly, countries have Systems in place offering a myriad of social benefits and services to meet the diverse needs of their populations. These programs can either operate in isolation supported by individual and program-specific information systems or work in a coordinated fashion (cross-functional or inter-sectoral), exploring synergies in design as well as reducing public and private costs associated to common phases of the delivery chain through streamlining and harmonization.6 13. Many countries use technology to establish information systems (IS) that facilitate the collection, storage, and analysis of data for single or multiple programs, and to support the delivery and management of these programs. The IS provides support in identifying potential program beneficiaries, facilitating their selection and enrollment into the programs, 5 Lowe et al. (2023) shows how the use of new technologies for supporting the delivery of social protection programs but emphasized that pace of adoption has varied between countries, despite its acceleration since the onset of the COVID-19 pandemic. 6 GIZ (2019) and Leite et al. (2017) 5 managing payment transactions, conducting monitoring and evaluation, case management, and generating reports. Therefore, IS improves efficiency for both people and administrators. 14. For people, the IS can help access more benefits and services, as it allows individuals to apply for several programs simultaneously through simpler procedures, leading to cost reduction and the provision of public services in a more efficient and coordinated manner. For administrators, IS improves efficiency by reducing staff time for processing information, improving data quality, governance, and transparency, and lowering administrative costs for processing transactions such as facilitating intermediation and referrals. 15. However, IS differs significantly from country to country and even from program to program. Some countries operate multiple, parallel, program-specific monitoring IS (PMIS) to collect, store, and analyze data for specific programs. Others have more complex data systems, integrating PMIS with central registries. For example, some countries employ integrated and interoperable social registries that contain information on potential beneficiaries for multiple programs, alongside integrated and interoperable beneficiary registries containing information on actual program beneficiaries. 16. Countries leverage technology to expedite communication, outreach, intake and registration, assessment of needs and condition, benefit delivery, access to services, payment systems, monitoring, among others, hence help enhancing administrative efficiency and transparency, and fostering national cohesion of multiple interventions. 17. Leveraging new communication channels through social media, emails, text messages (SMS), internet ads, website notifications, etc., can increase the quality of outreach for social protection programs. For example, during the COVID-19 pandemic in Bhutan, the Druk Gyalpo’s Relief Kidu cash transfer scheme received widespread publicity through television, radio, press, and social media. Similarly, in Colombia, television and social media played pivotal roles in disseminating information about the new cash transfer scheme, Ingreso Solidario. In the European Union, countries have ramped up their outreach efforts through social media platforms like Facebook, Twitter, LinkedIn, and YouTube. 7 18. This communication also enables provision of timely reminders to beneficiaries, and consequently improving other outcomes. These communication channels can inform beneficiaries about payment schedules and co-responsibilities, fostering greater compliance and engagement. Furthermore, the establishment of grievance and redress mechanisms 7 Lowe et al. (2023) 6 facilitated through toll-free hotlines, emails, and SMS ensures beneficiaries can voice concerns and seek solutions promptly. 19. Leveraging digital platform designed to operate in offline settings and synchronize data upon connection to the internet can serve as a channel for collecting and verifying necessary applicant information applicants while ensuring interoperability with other administrative IS such as civil registration, foundational ID systems, or unconventional early warning systems. The interoperability allows keeping information linked, up to date, support operations, and boost capacity for consistency checks and secure cross-check and cross- verification, which helps reducing errors, through automated with these other administrative data sources. This expedites intake and registration and processing times, minimizing the need for extensive data collection during intake and registration processes. 20. Moreover, technology enhances the assessment of needs and conditions, strengthening the capability to determine population eligibility for specific programs or across multiple programs, and consequently improving the accuracy and fairness of eligibility decisions. 8 It also allows better understanding of the diverse needs and vulnerabilities of the populations covered, and monitoring, by social programs using data analytics and tools. Data from various sources (e.g., household surveys, administrative records, or satellite data) are harnessed to profile pertinent populations, run dashboards, chart the coverage of social programs, map out service availability and delivery, assess population needs and vulnerabilities, estimate the likelihood of shocks, assess the quality and delivery of benefits, and track the frequency of shocks, as well as the progress and impact of programs over time. 21. Technology is also employed to facilitate payment transactions directly to beneficiaries through electronic payment systems, such as mobile money, digital wallets, or debit cards, particularly in areas where e-banking systems are functional; as well as access to certain services. 9 This approach can not only reduce the costs and inefficiencies associated with traditional payment methods, such as cash or check payments, but also improves the speed and convenience of payment delivery in areas with better connectivity. 22. In some countries, various strategies are adopted: some implement direct bank transfers to beneficiaries’ accounts (e.g., Brazil, Pakistan and Nepal); while others rely on mobile money accounts/digital wallets (e.g., Colombia and Kenya); as well as on debit/pre- 8 Aiken and Ohlenburg (2023) 9 Lowe et al. (2023) for a broader and detailed discussion about technology use for the delivery of benefits and services. 7 paid cards (e.g., Chile response to shocks) or e-vouchers (e.g., programs supported by World Food Program). For countries (such as Chile, Italy and Greece), technology is improving case management for social workers and improving referral services. 10 23. During COVID-19, countries such as Peru leveraged technology for remote visits and gathering information from beneficiary families, activities typically conducted in-person during normal times, ensuring families were properly informed about the COVID measures. Moreover, digital delivery mechanisms and channels can also enhance inclusion and access for vulnerable household members, such as women, in certain contexts. 24. Technology can also enhance preparedness and facilitate post-shock access to social programs. With economic and natural shocks, and disease outbreaks becoming more frequent, past gains in poverty reduction and human capital acceleration face the risk of setbacks. A robust system plays a pivotal role in mitigating the impacts of shocks while increasing resilience against future shocks. Studies by Bowen et al. (2020), the World Bank (2023), UNICEF (2019, 2021), and the Red Cross (2017) underscore the potential of technology in enhancing data quality and management to ensure timely and effective responses through ‘shock-responsive’ systems. 25. Technology-enabled solutions include online damage assessment (e.g., Chile) 11 and new online application platforms for affected populations (e.g., deployed in Brazil during COVID-19 pandemic)12 that are integrated and interoperable with existing social registries. Furthermore, leveraging Big-Data - public or proprietary data sets that are too large or complex for traditional data-processing application software - has allowed for identification of affected areas or individuals, as evidence in Togo, Democratic Republic of Congo).13 Interoperability with early warning systems enabled prediction of shocks before they occur, as seen in Benin.14 Additionality, technology expedites the secure transfer of benefits electronically, exemplified in Pakistan as part of 2022 floods response. 26. Technology enables better, faster, more efficient, and effective management and administration of social protection programs. With the more data available, the establishment of direct communication channels with implementers, and robust interoperability mechanisms ensuring data privacy and security, administrators can 10 World Bank Case Compass website for more information. https://www.case-compass.org 11 the Ficha Basica De Emergencia (FIBE): http://sise.gob.cl/ or https://www.chileatiende.gob.cl/fichas/89601- ficha-basica-de-emergencia-fibe 12 World Bank (2021) 13 Okamura, Ohlenburg and Tesliuc (2024) 14 Red Cross (2017) and Ijjasz-Vasquez, Jongman and Suarez (2017) 8 streamline monitoring and evaluation activities while ensuring transparency in decision- making and information dissemination. A myriad of initiatives that includes development of digital dashboards and maps for enhanced visualization and tracking of benefits and services and to generate regular reports that provide high-level management with valuable insights into program implementation progress and effectiveness. 27. In summary, the use of technology in social protection is likely to continue expanding and evolving in the coming years, with an increasing number of countries and organizations acknowledging its potential to enhance the efficiency, effectiveness, and overall impact of social protection programs. Moreover, the integration of ethical and responsible Artificial Intelligence (AI), which includes using Machine Learning (ML) models, principles into social protection information systems holds the promise of further enhancing efficiency and effectiveness of these systems and utilization of data. III. The risks and challenges of technology in the System A. An overview 28. This section delves into the risks associated with digitization and the broader use of technology by contextualizing the discussion within a comprehensive framework. It expands the lens of analysis to include both external factors influencing change and internal principles and components of digitization that shape efforts to digitize or leverage technology. This includes institutional arrangements, political environments, and legal dimensions, along with technical issues aimed at improving social inclusion rather than generating exclusion. We acknowledge that these risks are not exhaustive, as we do not extensively discuss issues related to country IT infrastructure, potential vendor lock-ins in contracts, or the need for system performance metrics, among others. These system design features are covered in other publications such as Linder et al. (2020) and Karippacheril et al. (2024). 29. As technology continues to develop and innovate (with new capacities, new data sources, and new computing power) and becomes more accessible than ever to individuals (due to the massive penetration of internet and telecom companies), it can become a transformative tool to help systems better deliver services to the vulnerable. However, one must be cautious about its use, risks, and challenges since technology is not an end in itself. For example, if a process contains a choke point or a glitch, or if data is not of good quality or exhaustive for certain groups of interest, then the automation of that same process will inherit that flaw when unnoticed, potentially generating exclusion. Hence, technology should 9 be seen as a means to improve the quality and efficiency of the established process, not as a solution per se or a substitute for human resources. 30. Grosh et al. (2022) highlight that program outcomes are as good as their objectives and processes, with sound functioning and rule-based behavior being clear and well- designed, which technology may help implement. Lowe et al. (2023), Lindert et al. (2020), and GIZ (2019) suggest that the transformative aspects of the technology revolution can significantly improve certain aspects of the delivery chain, making people's lives better. Therefore, it is important to understand people's experiences throughout the value chain, identify the problems that would make the experience difficult, and determine how technology can help address them. While some problems may be addressed with IT solutions, others may not. For example, if a mobile or web-based application for a social protection program is only accessible through the internet, how would populations without internet access be able to benefit from the program? Similarly, if only electronic payment is available, how would beneficiaries without access to their national identity documents or the e- banking system benefit from digital cash? 31. Moreover, from the administrative side, technology can improve transparency and accuracy, but determining the gaps that technology needs to address is equally relevant to this discussion. Often, interventions are designed without a clear articulation of the problem (also known as the policy cycle) and without mapping the required technological tools (and their intended outcomes) to address the problem or challenge. For example, is the problem solely about needing support for handling a large volume of transactions and data, about better transparency and informed communication between citizens and service providers, or about managing error, fraud, and corruption? Grosh et al. (2022) emphasize the importance of continuous investment in improving program outcomes in general by stating that "secular changes in technology do not automatically ensure progress," suggesting that good outcomes are always the result of political will, well-rounded administrative effort, and sufficiently resourced initiatives needed for sound social policy outcomes. 32. Technology requires, among other things, proper institutional arrangements, a conducive political environment, and regulations for data protection and exchange between multiple stakeholders. Political will and clarity on the roles and responsibilities of multiple institutions and stakeholders involved are needed to ensure data quality across different administrative systems and to bridge the digital divide or digital illiteracy among intended populations. Lowe et al. (2023) reinforce this by suggesting that the desired impact of 10 technology relies on strong legal and institutional frameworks, digital and financial services infrastructure, digital government capacity, digital access and literacy, and political will. 33. The rationale behind these points is that technology does not single-handedly fix problems without linking them to programmatic objectives or outcomes and transforming how programs function and deliver. For example, without political will, technology does not generate coordination, because main actors and stakeholders need to work together. Electronic data collection does not automatically mean quality data if the "interviewer" is not properly trained on how to extract and record the correct information. Sophisticated case management systems will not link individuals to services if the caseworker does not understand their functioning and is not aware of all available services. 34. An effective system relies on robust programs, proper institutional arrangements, transparency, strong partnerships, accountability, clear roles and responsibilities, and up-to- date data and robust information systems. By leveraging technology, coordination can be enhanced, allowing for the leapfrogging of traditional (and possibly slower) bureaucratic practices and the more effective delivery of services and system integration. In other words, the appropriate use of technology builds a bridge between the people and their institutions, improving the experiences of both the populations and the administrators of the system. 35. More specifically: a. For the population, technology has the potential to reduce transaction and private costs (such as the need to travel or miss days at work, etc.). b. For the administrators, technology reduces heavy administrative burdens, costs, and the duplication of processes. It can help profile the evolving needs and conditions of populations, identify who benefits from which programs, thus reducing gaps and duplications in coverage, and create synergies across bundled services, especially when coupled with ongoing efforts to enhance digital literacy. 15 36. However, technology have been oversold as they bring significant risks to the system and to the population in need. Most of the risks are often ignored or hidden between the lines when policymakers are promoting solutions to address system gaps. Nevertheless, many of the risks can be predicted and minimized to avoid unintended surprises. 15 Leite and Karippacheril (2019) 11 B. A brief technical dive into risks 37. Risks are often identified during the practical implementation but should have been identified during planning phased. The risks are higher during implementation when policymakers do not (a) consider the potential negative impacts of the technology on the population, and (b) promote the necessary investments in human resources as technology does not replace the need for more and better human resources in the delivery information system (or delivery system), as interactions between individuals and institutions occur nonstop on a daily basis. For example, countries such as Brazil, Chile, and Türkiye exemplify the importance of investing beyond technological advancements. These countries benefit from continuous investments in better planning, improved processes, and more equipped and trained human resources to address the needs of their populations. Such systems can be expanded (and contracted back) in times of shocks to support emergencies. 38. Institutional and Governance risks are significant and associated with having the proper institutional arrangements, political environment, and legal frameworks needed to ensure that technology improves data quality, data privacy, protection, and coordination. Expanding the use of technology can improve data quality and coordination among stakeholders only if each stakeholder is onboard and ready to share data via integration. Strong data privacy and protection regulations must be in place to ensure good data governance and protect individuals' information. Every stakeholder, including third-party actors such as Non-Governmental Organizations and development partners, must be committed to participating. For example, policymakers often assume that one agency in charge of the system can simply request data from another institution, which is not always the case. 39. Due to the manipulation of large volumes of information about individuals' socio- economic conditions, data privacy, protection, and training/skilling activities for the team working in such a system are essential. Even with AI and Big Data, training/skilling and in- house human resources capacity are crucial for developing algorithms and models to ensure accuracy and transparency. 40. Lack of training/skilling for people in charge of registration and case management, or for running AI processes and models, increases the risks associated with misuse or improper use of information, compromising data privacy and security. For example, despite the growth of social registries and case management, many still treat the interaction between agents and individuals as a data collection exercise rather than a service provision. Properly designed processes and trained human resources are essential for ensuring data quality and effective 12 service delivery as such interactions of the government with the individuals to provide a benefit or a social service are the real representation of a good program and of treating individuals as humans, and AI models alone are insufficient to address this need of human interactions. People must be able to interact and understand the constraints and needs of the population to provide the necessary support, and consequently have the appropriated training and skills. 41. Technical risks include issues related to data systems, such as inconsistency of information, timeliness of data, different units of reference, and data dictionary inconsistencies. These issues increase the chances of unsuccessful data integration and system interoperability. Multiple agencies may collect data at different times, for different units of interest, and with varying protocols and specifications. For example, income tax systems may be updated annually, property taxes only when a new property is registered, and civil registration only when an event is declared. Simple data fields as age that can be collected as MM/DD/YY versus MM/YYYY or DD/MM/YYYY; or name that can be collected as first, middle, and last versus first and last, and protocols allows for abbreviations in one but not in another system. Such inconsistencies can lead to data integration challenges and should not be placed on the shoulders of the applicant. Integration and interoperability are crucial for minimal data collection but without addressing such inconsistencies’ they simply do not occur. 42. Due to multiple actors using the system, there is a significant risk of information misuse. Proper accessibility and control protocols must be enforced to protect data. In a blink of an eye, someone could access the full set of information, from ID to bank accounts, of an individual in the system. Proper safeguards, such as secure authentication and firewalls, are then essential. 43. Another technical risk is associated with the need for individuals to provide consent for data disclosure and sharing. Without proper consent forms and data management, unauthorized use of information can lead to lawsuits and loss of credibility. This risk is heightened with the growing use of Big Data and AI algorithms, which can introduce systematic biases and increase ethical risks. 16 44. Risks related to digital access, coverage, and literacy must be considered. The goal of using technology is to facilitate individuals' access to the system. Without understanding or assessing people's needs and experiences, technology can penalize rather than improve their experience, generating exclusion. For example, if only a mobile or web-based application is 16 United Nations Systems (2022) 13 available, populations without internet access may be excluded. If the application is data- heavy, the poor may not afford it. If the application requires extensive information, digitally illiterate individuals may struggle to use it. Similarly, if only electronic payment is available, beneficiaries without access to national identity documents or e-banking may be excluded. And in the context of recurrent climate-shocks, the reliability of technology and protection of System are critical risks. System failures, for example due to power outages, can occur and data can be simple erased, impacting stakeholders. Robust contingency plans and resilient infrastructure, including having backups, are necessary for effective use of the System, including to respond to shocks. 45. In sum, numerous risks must be considered and addressed when using technologies. Policymakers must identify these risks and develop tailored solutions for their countries. Technology is an enabler that facilitates interactions between the government and the population. Its utilization should be based on proper institutional assessments and designed to protect the population, supporting a robust system. The next sections will expand on these challenges and opportunities and provide strategies for addressing many of the risks to enhance the system. IV. Addressing these risks A. Institutional Arrangements 46. The multitude of actors and complementary programs aligned with SP objectives underscores the necessity for effective leadership, a well-defined legal framework, and clear delineation of boundaries and institutional arrangements. These elements are crucial for anchoring the planning, management, and delivery of assistance, particularly as it is intertwined with and supported by the broader use of technology. 47. Government leadership assumes a central role in coordinating diverse actors and instilling the importance of budgetary allocations among key actors. The establishment of a robust legal framework is essential for ensuring the proper management and utilization of required data. Meanwhile, institutional arrangements play a vital role in defining the respective roles and responsibilities of key actors. 48. Considering the political economy, attention to legal or administrative and institutional aspects becomes even more important. As the System relies on network connectivity (see more on institutional factors and digital exclusion), it is important to 14 consider involving other government-wide agencies responsible for providing digital infrastructure across the country. This ensures that data governance, data infrastructure, and safeguards, which are critical elements of the System, are in place or can be developed. 49. Coordination of various actors prompts the government to effectively organize information flow, define roles and responsibilities, and ensure transparency in delivering benefits and services to the population. Leveraging technology presents an opportunity to enhance coordination. However, diverse stakeholders may have individual incentives to operate their systems independently, including humanitarian systems, resulting in inefficiencies. Therefore, it is essential to consistently address the coordination challenge across central and local institutions, as well as between government programs and humanitarian actors. Providing incentives for collaboration that benefit all parties, and the collective can be a strategic approach. 50. For example, in Mali, during the inception of the Jigisemejiri 17 program after the 2012 coup d’etat, the government took proactive measures to prevent political capture and garner collaboration from various sectors. To ensure an intersectoral approach, Jigisemejiri was set within the Ministry of Economy, Finance, and Budget (MEFB). To institutionalize its operations, a National Steering Committee (NSC) was established through a Prime Minister decree (#2013-0195, February 27, 2013). The NSC's primary objective was to provide policy guidance, with its members drawn from relevant sectoral ministries. The NSC played a pivotal role in setting policy goals, approving annual work plans and budgets, reviewing progress reports, and monitoring the impacts of Jigisemejiri activities. Additionally, the NSC supported the establishment of the Unified Social Registry (Registre Social Unifie) for Mali. As of March 2022, the NSC had organized 18 sessions, consistently chaired by the MEFB. This structure facilitated the Jigisemejiri program's continuous operations amidst numerous challenges. It enabled the program to provide access to regular cash transfers, labor-intensive public works, income-generating activities health insurance (RAMED), and emergency cash transfer program linked to COVID-19, all in a coordinated manner. Furthermore, Jigisemejiri partnered with the ’Commissariat à la Sécurité Alimentaire’ to ensure coordinated and efficient interventions addressing food security. 18 17 This program was the first step towards developing a social assistance system in the country. More at https://www.jigisemejiri.org/ 18 RAMED - Medical Insurance Scheme for the Vulnerable - was expanded in 2018 through the collaboration of ANAM and Jigisemejiri that reduced registration burden as RAMED was offered to previously identified Jigisemejiri beneficiaries and those enrolled in the RSU. 15 51. It is also necessary to improve intersectoral government collaboration (e.g., communications and dialogue, and funding) to promote integration and interoperability. Establishing a collaborative model among key actors is equally essential to ensure their responsibility for data and information, facilitating proper processes for data-exchange and service provision. (e.g., combined outreach and information campaigns or enrollment in social registries) that would deal with cases of data inconsistency between different administrative sources. For instance, in Chile, the monthly updating of administrative data is meticulously processed. However, each partner institution within the government determines which data to extract. The monthly update aligns with the periodicity and temporality of other administrative data. Some data may remain unchanged for 6 or 12 months, reflecting the administrative data recertification period of either the semester (Indigenous registries) or the year (tax revenues). 19 52. The main goal is to ensure that there is no data-exchange resistance from stakeholders, so the agency in charge can effectively provide benefits and services. However, data-exchange can be affected by the ownership of sensitive data, such as tax-related information (payroll for firms, sales for VAT, personal income, and property details like land or automobiles), collected for specific purposes. The need of revising legal aspects to access this type of data will be discussed later. 53. Additionally, since connectivity results from investment in multiple sectors, it is essential to involve agencies responsible for providing basic infrastructure and internet services when developing a System. This collaboration can help expand the System to areas likely to be excluded from the grid due to low demand, poor services, and high costs. Discussing the System on a broader scale can help identify alternative solutions or economies of scale to increase network coverage. 54. In summary, effective SP systems require institutional collaboration among key stakeholders that enable successful navigation of diverse policies and strategies across different sectors as exemplified here by Chile and Moldova: a. in Chile, a country with high vulnerability to natural disasters, the integration of social protection and disaster risk management was strengthened in 2015. The overall disaster risk management system was revised to be more modern, precise, and faster, focusing on aiding families with social vulnerabilities, providing better training for response teams, and streamlining public 19 Table 3 in Silva et al. (2018) 16 communications. Accordingly, the Emergency Basic Fact Sheet (Ficha Básica de Emergencia, or FIBE), along with mobile apps and tools developed, was made interoperable with the Chilean National Social Registry. FIBE electronically collected on-site information of affected so that public servants could refer the population to available benefits and services. As a metric of efficiency, data collection in response to the Coquimbo earthquake in 2015 took 27 days using the new FIBE, contrasting with the response to the Tarapacá earthquake in 2014, which took 115 days, in part due to the interoperability with the Chilean National Social Registry. 20 b. In Moldova, the Moldovan Social Assistance Automated Information System (SAAIS) was developed to enhance the efficiency of social assistance programs. While the development of SAAIS benefited from a sound legal and institutional framework, the finalization of service agreements between agencies for data exchange proved to be a cumbersome process. Nevertheless, the system today exhibits robust interoperability, allowing social workers to input a citizen's application, with the system automatically retrieving data from various agencies through web services. 21 B. Political Environment 55. Political will encourages or mandates the alignment of SP programs toward a broader goal. Political determination helps SP to be acknowledged by legislators, taxpayers and the general population and recognize SP as a part of the social contract - a policy to help protect against and mitigate poverty and reduce inequality. 56. Political will also promotes the alignment of sector specific visions that are compatible to general social protection goals – which includes, among others, reducing poverty, promoting rights, increasing human capital, and boosting productivity – thereby creating the political environment needed to have sufficient resources and grit by stakeholders in order 20 This integration of FIBE with the Chilean National Social Registry significantly reduced data collection time, as most basic information for affected families is prepopulated in the FIBE information system. Moreover, through this organized approach to addressing shocks, the centralized platform enables all relevant institutions to access real-time data, communicate pertinent information to the public, and serves as a gateway for citizens to contact the government directly as needed. 21 The SAAIS includes data from the Population Registry (applicant and family members), National Transport Registry (registered vehicles), Employment Agency (employment status, information on unemployment benefits, and job rejections), National Office of Social Insurance (pensions and benefits), National Cadastre Registry (land and immovable property), and Border Guard Service (to determine the beneficiary's current location). Additionally, the system conducts automatic monthly reconciliation with all agencies before preparing payment lists. More in Sluchynskyy (2019). 17 to improve the quality of existing services, their transparency, efficiency, reach, and cost- effectiveness. 57. Political will comes the allocation of sufficient attention and resources to key stakeholders, to run a consolidated delivery system that is well-funded, staffed, dynamic, agile, and ready to support vulnerable people. Political will tailors the agenda by pushing for coordination and by providing the financial support needed to boost technology gains. Without such focused investments, technology is not enough to generate transformative results. 58. It is important to observe that as countries unite in establishing consensus on transitioning towards Universal Social Protection, Adaptive Social Protection, among other global agendas, the role of political will becomes increasingly important. Governments are tasked with the delicate balancing act of allocating resources and budgetary considerations to Social Protection (SP) programs. These programs may encompass targeted benefits for specific and diverse groups, alongside other sectoral initiatives in areas like Education and Health. The challenge lies in enhancing the efficiency gains of these programs while maintaining a strategic equilibrium in resource allocation. 59. To highlight the political will importance, we describe below the experience of Türkiye for the development of their Integrated Social Assistance Service System. The System was built to improve the efficiency and timeliness of application for social assistance programs that includes family benefits, conditional cash transfers for improved health and education, old-age pensions, disability benefits and care services. Until 2009, applications for social assistance programs were entirely paper-based, and each social assistance program had its own process. Moreover, citizens had to obtain documents from various organizations to verify their information. Through the creation of unique service windows and ensuring the interoperability of administrative systems which were built between 2010 and 2015, social assistance beneficiaries can now apply for benefits using one document- their national ID card. The System links with other institutions, such as the Ministry of Family and Social Policies, Ministry of Interior, Ministry of Labor and Social Security, Ministry of Finance, Ministry of Health, and Ministry of National Education, and such a transformation made the processes electronic, generating public and private savings such as in processing time (e.g., from about 15-20 days to less than a minute for the initial verification of eligibility), reducing paperwork (e.g., processes approximately 2.3 million fewer documents per month) and staff time (e.g., from application to decision dropped by approximately 20 percent). Consequently, 18 this approach generated efficiencies as the time required from application to disbursement of benefits (of the disability and old-age pension programs) dropped from 1.5 years to one month. The System development was backed by strong political will as the System consolidates information from multiple agency systems. Its development was initiated with the Deputy Prime Minister, which helped provide the impetus for concluding key data sharing agreements and partnerships across multiple institutions. As part of the process, special attention was given to the communication about the value of data sharing as government institutions were sometimes reluctant to share data or put the effort in to digitizing databases, which resulted in significant time savings and efficiency for many ministries because they no longer had to process requests for documents for social assistance applications. And, because integrated information systems are complex and continually evolving, Türkiye’s government paid extra attention to recruit and retain skilled staff members, providing regular capacity building at central and local levels, and conducting trainings both online and in person. 22 C. Legal dimension 60. The term "legal" or regulation, referred to in this section and in the context of technology carries diverse meanings. With numerous institutions involved in supporting benefit and service provision, data exchange, privacy, protection issues, and susceptibility to cybercrime and bad-intended manipulation arise due to extensive manipulation of data. Hence, a comprehensive review of the current regulatory framework for data exchange, privacy, and security within the country. 23 61. Regulations are then imperative due to SP's management of personal data, encompassing any information related to an identified or identifiable individual. Additionally, it deals with personally perceptible information, which includes data permitting direct or indirect inference of an individual's identity and related data points, such as street address, email, telephone IP address, geolocation, biometric, and behavioral data. This sensitive data can also include information related to sexual orientation, membership in an ethnic or minority group, or affiliation with a trade union, which can raise more privacy concerns. Hence, proper governance and regulation for data privacy and protection must be in place, and data security protocols, including different levels of access for data manipulation, 22 Turkey’s Integrated Social Assistance System Report. 23 Alston (2019), GIZ (2020), Sepulveda Carmona (2018), Bashir et al. (2021), and UN Personal Data Protection and Privacy Principles. 19 firewalls and robust and secure backups are necessary to ensure that data is protected, secure and safe. Moreover, in general sensitive data such as tax-related information (payroll for firms, sales for VAT, personal income, and property details like land or automobiles) are collected for specific purposes. To utilize these datasets for government-provided services, new laws must clearly define the permissible capacities in which such systems can be employed and address the associated legal aspects. 62. Determining the feasibility and establishing the necessary arrangements for government access to such sensitive data involves considerable challenges, navigating which requires a robust combination of significant political will and a well-defined legal risk/framework. Data-exchange of such data across agencies is also not straightforward due to data regulations, and having services agreements between agencies for data exchange, which can be a very cumbersome process, and one may not avoid some legal revision of current regulations and creation of a new institutional framework to allow such data exchange. 24 63. This entails a thorough examination of existing privacy laws and regulations that govern specific data, considering the need for consent from individuals as part of the interoperability process for administrative purposes. For example, in Chile as part of its system modernization the Government by law 25 created the Ministry of Social Development (Ministerio de Desarrollo Social) and clearly specified its functions and attributions including access to tax revenue data and defines the roles and responsibilities of various institutions in handling sensitive data. 26. Furthermore, the right to privacy for all individuals is recognized, protected, and guaranteed by the Chilean Constitution (Article 19), with a 2018 amendment explicitly establishing the protection of personal data as a constitutional right. 64. Institutions such as the World Bank, GIZ, WFP and UNICEF have been vocal on the importance of protecting data for social protection. However, there is still a long way to go to ensure that good privacy and data protection laws are in place in many low- and middle- income countries. One approach to tackle such data privacy and protection issues, especially in the absence of a robust legal framework within a country, is to consider the European 24 Data-exchange importance increases in a post-shock response, as rapid response could lead to exclusion in case people are not registered in any existent system. Some agencies started looking at this issue such as the World Food Program that has developed a guide titled Social registries, Data processing & Appeals first produced in 2021 at the midst of the COVID-19 pandemic, during which governments were actively asking WFP and other actors for support in developing and updating their social protection registries, and in beneficiary data exchanges for a rapid emergency response while ensuring people have mechanisms to appeal when excluded. 25 Ley 20530 at https://www.leychile.cl/navegar?idNorma=1030861 26 Silva et al. (2018) 20 Union's General Data Protection Regulation (GDPR), which can serve as a valuable reference point for developing local data protection regulations tailored to SP programs. Collaborating with the government is important in this process, ensuring that the privacy of individuals participating in the evolving System remains a central focus. For example, in 2022, the Social Protection Inter-Agency Cooperation Board (SPIAC-B) published a report focusing on good practices for data protection and data privacy, stressing the need for security and respect for the right to privacy and ensure the protection of personal data. 27 65. Moreover, as Big Data and AI gain popularity it becomes imperative to establish legal protections concerning data classification, ownership, and protection. The rapid expansion of Big Data (e.g., data derived from sources like satellite imagery, mobile operators, web, and social media) alongside with AI (e.g., machine learning techniques) necessitates caution due to the private nature of much of this data. Individuals often have not given consent for their information to be used for various purposes. Given that many Big Data sets are privately held, and there is an ongoing process to determine regulations or incentives for making such data available to the public to support government functions, promoting the use of such data requires careful consideration. Furthermore, the regulatory landscape for Big Data, particularly concerning sources like mobile phone data, is unique to each country, and data access constraints vary accordingly. Consequently, the use of big data in one country does not imply that the same data can be seamlessly employed in another country. Policymakers must, therefore, address fundamental questions from the outset: What regulations or incentives should be in place to ensure the continuous availability of such data for essential government functions, extending beyond crisis situations? Additionally, what is considered socially acceptable in terms of government access and for what specific purposes? 66. To illustrate this case, we highlight the Togo experience. In Togo, the Novissi program leveraged big data for identifying potential beneficiaries during the COVID-19 emergency. However, accessing this data posed significant challenges due to legal and regulatory bottlenecks. Mobile data collected from mobile network operators (MNO) could not be directly utilized by program administrators due to existing legal requirements on data protection aligned with both National and International standards. This necessitated the signing of non-disclosure agreements between MNOs and the government, and, more 27 The SPIAC-B is composed of members from United Nations system organizations and bilateral development agencies, donor governments, and civil society organizations. More on data protection in SPIAC-B (2022). 21 importantly, data access and processing had to be conducted with the explicit consent of subscribers. 67. As inter-agency coordination is needed, legal mandates and official collaboration will push for each stakeholder to “cooperate”, setting the formal cooperation agreements, defining budget-sharing, or another administrative-cost-sharing arrangement so that everyone knows their roles and responsibilities. This inter-coordination is important for multiple task such as compliance monitoring of conditionalities that can be complex as the number and diversity of actors (e.g., health and education sector agents in conditional cash transfer for children and adolescents) involved increases. One legal issue that needs to be addressed to ensure that key agents (such as frontline program personnel, teachers, health care workers, subnational agencies, and central agencies, including the social, education, and health ministries) can perform the tasks needed is related to the implication of the new tasks into their existent work program. Some of the people can be already overburdened and tasks required would be considered additional not previously agreed in their terms of reference. Ensuring that all stakeholders are part of the system and do not see social protection needs as additional work requires proper institutional communication and the set-up of legal boundaries between stakeholders. 68. Setting up inter-sectoral committees can help delineating the work and boost coordination and obtaining support from high levels of government to streamline social protection in the country. One good example that tries to legally address this issue is Pakistan. In the province of Sindh in Pakistan the newly established Sindh Social Protection Authority is the lead social protection agency. For effective coordination between the Sindh Social Protection Authority and other implementing partners and departments, and to better integrate systems, including with Benazir Income Support Programme (BISP) at federal level, and generate synergies between multiple interventions that aims to improve living conditions and resilience of poor population in the province, the steering committee membership comprises the CEO of the Sindh Social Protection Authority, along with the Secretaries of Finance, Health, and Social Protection. Other members include the representative of health service facilities administration, head of Social Mobilization partner and two members from non-governmental organizations active in Social Protection. The committee also invites BISP to participate in their meetings for better provincial and federal coordination. Moreover, as follow up of the first National Conference on Social Protection organized by BISP with financial and technical support of World Bank, GIZ, ADB and UNICEF 22 held in May 2023 28, a Coordination Committee lead by the Ministry of Poverty Alleviation and Social Safety was expected to be formed, and one main responsibility of this committee was to have focused discussions between the federal and provincial level, in particular regarding the full interoperability of data system for efficient data sharing between the National Socio-Economic Registry (NSER) and provincial data. 69. Addressing the required interoperability of systems effectively requires the synchronization and harmonization of basic standards to ensure that the jurisdictions, competencies and responsibilities of the participating stakeholders are well defined. Implementing regulations that advocate for standardized operating procedures (SOP) from the outset or during development is key. Even if a system begins with a single user program, it should be conceptualized with a vision of future interoperability, which involves configuring/enabling APIs during development. One good example is Nepal that has implemented the Nepal e-Government Interoperability Framework (NeGIF), providing a framework for sharing, collaborating, and integrating information and organizational processes using common standards. The NeGIF outlines standards for Government-to- Government (G2G), Government-to-Citizens (G2C), Government-to-Businesses, and Government-to-Employees interactions. As currently, the Department of National ID and Civil registration started working on internal data sharing between National ID and Civil Registration systems, and the development of the integrated social registry, guidelines and SOPs are being developed following and guidelines are aligned with the standards set in the NeGIF as the work deals with G2G and G2C categories. 70. In summary, government, private sector, institutions and programs must first be sure about the existent legal environment and legal framework for e-government development, then they must comply with laws governing information such as, to name few, personal data protection, digital signatures, interoperability, and information security. D. Data systems29 71. Data and digital Management Information Systems (MIS) are central to Systems. They play a crucial role in data collection, determining eligibility, and managing programs throughout the entire value chain, including benefit delivery. Data is not instantly produced 28 The first annual National Social Protection Conference in Pakistan brought together stakeholders from all over the country and international experts to discuss how to develop a System that is better equipped to address the needs of the most vulnerable, particularly in the wake of climate change https://www.giz.de/en/worldwide/125090.html 29 Grosh et al. (2022) 23 and depends not solely on technology. Effective data management requires specialized skills to run a seamless system, combining specific human resources with technology to collect and assemble accurate data sets, storing them in a structured manner for easy and secure use and the transformation of data into valuable information and insights. Data originates from information collected from new applicants or existing datasets accessible to the government and other stakeholders. 72. To establish efficient information points, SP utilizes a secure and trusted architecture for data exchange among ministries and agencies. This involves collecting and digitizing information from all business functions, implementing data quality control routines, and integrating data subsystems from each business function to allow automatic detection of data gaps, incompatibilities, outliers, or changes that can trigger adjustments in other systems affecting the broader system. Yet, several technical challenges must be considered in Data management. Factors such as data security, privacy, protection, availability, recency, reliability, identity, and lifecycle management (including collection, storage, retrieval, manipulation, augmentation, archiving, and deletion) are critical when designing systems, processes, procedures, standards, and governance for registries, systems, and related data dictionaries and types (meta, micro, structured, unstructured, organic, synthetic, proprietary, static, dynamic, or open). 73. Inconsistent information from various sources is a challenge. Data, through interoperability, enables the exchange of information collected and updated at different points in time among various systems, leading to potential inconsistencies. Additionally, different units of observation may limit interoperability scope as a robust SP system should uniquely match individuals while simultaneously allowing the creation of household or family-level folders with unique identifiers as well: a. The timeliness of data poses challenges as administrative data management protocols varies among agencies, with different protocols for update and recertification. 30 Some systems are static, with infrequent updates (e.g., annual tax updates or census-style Social Registries, vehicle or property registration, etc.). In contrast, others are dynamic, such as social security, with monthly contributions and benefits payments, or the Guaranteed Minimum Income system. These systems may not necessarily cover the same time periods, creating a disparity in data currency. 31 30 GIZ (2023b) 31 Table 3 in Silva et al. (2018) 24 b. Diverse units of reference pose challenges, as most systems lack the capability to group individuals into households or families, with family or household folders being rare. This issue becomes even more pertinent with the availability of big data, where household or family-level information is seldom found. Even in geo-spatial analysis, the focus remains at a grid level rather than being household specific. Additionally, data from call records may be linked to the SIM or the phone number, which an individual or household may have separately or shared with other members, adding further complexity to the unit of reference. 74. Ensuring a unique "key" is essential for interoperability, which requires unique matching of individuals or households to integrate information. This reduces the need for redundant data collection, validates information and identity, and cross-verifies data. Digitizing foundational systems like National ID and Civil registration is crucial, and the extent of integration and interoperability of systems depends on the coverage and use of the national ID system. Social Security Numbers and voting IDs are also commonly utilized as unique identification numbers. While having widespread availability of unique numbers for everyone is vital, countries can employ alternative business processes for matching individuals or households across multiple systems. However, for this to succeed, systems must share a common architecture and data dictionaries. 75. More specifically on data dictionaries. Very often existent systems are designed independently without a vision for future interoperability. As a result, closed systems, especially those without support for interoperability and APIs (application program interfacing), hinder interoperability, primarily due to incompatible data dictionaries or structures. The absence of a standardized data dictionary across systems impedes interoperability, preventing the use of business processes for sharing information or matching individuals. Although the basic information collected is often similar, variables like name, date of birth, gender, education, and occupation are not consistently recorded in the same sequence or format across systems. Disparities also exist in date formats (e.g., MM/DD/YY versus MM/YYYY or DD/MM/YYYY), and codes for geographical areas, caste/ethnicity, and education levels are not uniform, making databases challenging to interoperate. To address this, a middleware can be developed to convert data formats, but the main action to be performed is to have an in-depth revision of IT architecture from all systems that are to be part of the System before hand to properly define an IT architecture and a work program that would promote changes if needed in existent systems to ensure that administrative databases can communicate effectively as part of the system. 25 Investments in the new and in the existent systems are needed for developing a harmonizing technology that language, coding, and standardized data dictionaries across multiple systems, which are all crucial when aiming for interoperability and data integration. E. Minimal data collection and access32 76. As integration and interoperability occur, users cannot have access to all information available nor the ability to change records without following a proper user protocol. The goal is to avoid the accessibility and control of non-required information to the administrative data holder control. Moreover, to prevent unauthorized access and control of unnecessary information by the administrative data holder, it is essential to keep the collection of identifiable information to a minimum for the sake of data protection and systems integrity. The final agreement between administrative bodies necessitates the preparation and signing of Memorandums of Understanding (MoUs) to ensure consensus at both technical and administrative levels among the involved parties. This is crucial due to concerns about the perceived lack of control over data, which may pose issues for the parties involved. 77. Systems must often share information to validate self-declared data in order to improve accuracy in the selection of beneficiaries. For example, when cross-verifying information with formal employment database, the System administrators do not necessarily need to know the value of the earnings but only if the individual works in the formal sector and have earnings above a given threshold. Or with regards to possession of goods, when sharing information of house or car ownership, the System just needs to know whether the goods is estimated to have a “value” above the threshold. Data administrators must retain their capabilities to control the quality of their data, and there should be awareness and agreement on how their data are reused by the System. 78. Hence from the technical side, the IT architect must create a system that does protect individual data from misuse by creating privacy layers internally, firewall protections externally, and varying audit levels of access and modification by different individuals that ensures the ability to identify changes and potential misuse of the system. Many organizations have developed standards and guidelines to better navigate the data topic, including governments, UN agencies, organizations, donors, partners and third parties. 32 OECD (2019) 26 F. Consent for disclosure and sharing of information 33 79. With technology, risks of violating personal, private, socially agreed, and contractual terms of data re-use increase. Consent, security, privacy and ethical considerations are essential to ensure that data is in the hands of the right individuals. Individuals must provide consent for both disclosure and sharing of information, since with information provided, privacy concerns arise. Consent must be provided by individuals at the time of registration in the Data systems in a free, specific, and unambiguous manner. 80. Moreover, individuals must be notified about the purposes, and that consent can be withdrawn at any time. Therefore, the system must be designed in a simple manner (e.g., having a website checkbox for providing and withdrawing consent, having statements that are easily accessed by user to read about the process and purposes of the data) to comply with the international standards such as European Union’s or national standards. For example, in Senegal, the National Unique Registry (Registre National Unique, RNU) protocols during the data collection phase include provision of information about the social registry, the potential users, the right to not respond (would prevent inclusion), and duration of data management. Then, households’ consent is requested before data collection begins. In Jamaica, for instance, health workers inform the users that they are the only ones to review beneficiary health records as part of the conditional cash transfer verification process. In Togo, while registering to the Novissi program, a consent form was included as a part of the registration process that also requested applicants to allow access to their data. 81. Data breaches, and the need for Safeguards such as security, privacy and back-up tools (e.g., firewalls, secure authentication, redundancy, etc.). As more Data is used by governments for transparency and efficiency, personal data breaches are at risk. Due to the increased volume of data breaches in today’s world Systems must have the proper safeguards that includes data encryptions and firewalls to safeguard personal data. Hackers are often trying to obtain personal informal for multiple purposes, and Systems can be a rich source of “revenues” for them as systems keep identification information, personal details (including cellphones) and bank account information in databases. Therefore, all proper security measures against potential attacks, malicious activities and accidents must be in place as personal data breaches and other digital security incidents like malware and phishing are likely to happen (some through suspicious email links). For instance, WFP piloted a 33 Lindert et al. (2020) 27 Blockchain technology, called ‘Building Blocks’, 34 authentication architecture to ensure security through the identity-protection and authentication of users, as well as access to SP and benefits delivery systems. This Blockchain technology is defined as ‘collection of blockchain hubs – computer servers that are independently operated by each participating organization. Together, they connect and allow humanitarian organizations a neutral space to coordinate the delivery of assistance.’ 82. Interoperability does not develop in a vacuum. There are several enablers and safeguards that facilitate inter-operability and trusted data sharing (Lindert et al, 2020). At one level, this concerns regulations and institutions that go well beyond what any particular social protection program has agency over: a policy and regulatory environment that define and enact rights over data; robust and resourced institutions capable of enforcing the rules while also offering citizens responsive and effective redress; technical architecture to standardize data sharing within government while giving people more controls and transparency into data flows; capabilities inside and alongside government to analyze and make use of data; and an active civil society and informed populace who can effectively use or let us their data and keep governments and companies accountable. 83. Another level of enablers comes from more technical investments that enable data sharing and data security: interoperable databases that are accessible to and used across government agencies for data sharing; e-services portals that allows citizens to access government services, individual data portals that allow people to aggregate, store, and share data, and inclusive digital platforms such as digital identification that ensure all people are participants in the digital economy. All these factors are crucial enabling or disenabling parts of the data ecosystem and influence how extensively an individual social protection program can use external info. G. AI and Big Data crescent 35 84. The Economist intelligence unit (EIU) address AI as ‘AI’s intelligence abilities when discussing potential and risk surrounding the topic. With all the attention and considerable resources invested in such a promising yet evolving concept, this section opens the discussion around challenges and prospects. ‘AI has made headlines with its cerebral power. A silicon chip signal is 100,000 times quicker than an electrical signal in the brain, which has shifted the fear of automation from blue-collar jobs (where robots were going to do all the work) to 34 WFP (2022) 35 Mukherjee et al. (2023), Okamura et al. (2024), and Grosh et al. (2022) 28 white-collar ones (where software will replace humans). Whether or not we get to artificial general intelligence (AGI), where a machine can do any task better than a human can, it appears that 2024 will see a lot more innovation, investment and regulation in different aspects of AI.36 85. However, AI is not necessarily looked at as a product or a solution, but rather an innovative design principle applied to contexts where data and systems can become more efficient, effective, adaptive, and intelligent, or even predictive in nature (applicable to anticipatory action in SP, EWER). These designs, allow technological systems to make data derived or automated decisions and predictions. Such a revolution seems to apply primarily to the systems engineering and the user-interface layers of engineering offering the potential to democratize the use of technological innovation to the masses, bypassing the initial capital investment cost of computing but albeit with limited consideration of the potential misuse of this innovation. 86. The promise of AI is solidified when statistical and functional models are designed with an adaptive and learning mechanism that enables them to adapt, learn, and progress with altered behavior and with limited human intervention. This principle is further advanced or accentuated when models are ‘fed’ with data sets to be trained and thereby can alter or enhance their functioning, decision making and sometimes behavior through analysis and understanding of incidents extrapolated from the data itself and fueled by sequenced algorithmic logic (a process or set of rules to be followed in calculations or other problem- solving operations). In the context of AI, Generative AI is a subset of AI capable of generating text, images or other data using generative models, often in response to queries or prompts. 87. Generative AI models can learn the patterns and structure of their input and training data. Such architecture has tremendous potential to improve current systems in ways that affect behavior and way of life, including in social protection. Furthermore, by assessing its applicability to social protection, AI has proven their ability to leverage digital innovation potential with improved targeting, data management, fewer exclusion errors, predictive modelling (Early warning, early response), better interoperability of systems, quality data sets (including open data), and more inclusive and transparent delivery of benefits through multi-channel delivery. 36 EIU ( 2024) 29 88. In addition, the concept of Open-source AI has been gaining traction, and therefore with the combination of digital public goods and a ‘responsible’ open-source AI there is potential for exponential advancement in the Digital/AI SP space. However, with this innovation comes several risks and challenges, some more severe than others, which are generally beyond the capacity of most individual organizations or governments to protect the vulnerable and mitigate, cope or adapt to the negative consequences of such change. 89. The question then becomes one of responsibility, transparency and the ethical 37 use of AI with a need for a referee that can monitor and guide the responsible functioning and implementation of such systems with an eye on risks and consequences. The known factors or consequences of applying such innovative AI technologies within a relatively short period of time and limited oversight include the following: Hallucination, bias, deceit, ‘fake news’, ambiguity, traceability, detection, obscurity, lack of privacy, security (possible breaches), ‘brute-force’ learning of models, digital dictatorship (despotism), foster surveillance, and data relevance. Data is an important piece of the design and historically, developed countries have amassed large amounts of data sets that can be used to train AI models – although with caution towards past bias - that developing countries do not readily have access to. 90. While the opportunities are seemingly promising and extraordinary, the risks are many, and while the pace of innovation and change has always accelerated through various stages of development, societies’ ability to prepare and cope has not necessarily kept up. It is therefore our responsibility to build the guidelines and recruit the referees to monitor such innovations and changes, ensure governments and non-governmental actors are transparent with their intentions, help people cope and adapt with their negative consequences, and ensure equity and transparency. 91. Global organizations including the UN and several national governments have stood up for the challenge. For instance, the US, UK, and the EU have developed several layers and frameworks of monitoring and accountability that balance regulation with innovation. One might argue that Data is the new ‘currency’ of the digital AI sphere, and the latest design principles models and products from AI certainly help strengthen this theory. According to the EIU (2024), “The debate between optimists and pessimists needs moderation, as AI can both have a positive impact and cause immediate risks”. 38 37 United Nations Systems (2022) 38 EIU (2024) 30 92. The transformative promise of technology in SP hardly takes place without considerable attention to change management. For instance, there are several factors that negatively affect digital transformation across the technical and behavioral dimensions. The technical challenges include dealing with legacy systems, handling talent shortages, managing limited budgets, security concerns, handling interoperability, dealing with data- related issues and addressing technological incompatibilities and architectural deficits. As for the behavioral challenges, they generally include resistance to change, risk averse culture, siloed decision making, lack of clear goals and objectives, inadequate communication, and lack of leadership and commitment. The tenants of change are rooted in people, processes or operations and tools. Designing a theory of change can be successful when attention is given to all three tenants of change. 93. Furthermore, technology on its own constitutes several dimensions of interrelated components that generally span from technical, functional, behavioral to data science factors. Combining traditional software development lifecycles with project management philosophies and adding ethical AI principles to the mix of ingredients, requires a blend of technical talent ranging from math, statistics, science, software engineering to deep machine learning principles and data analytics. 94. AI and Big Data have been presented as solutions to Systems, when they should be considered a tool to support it. For example, recently during the COVID-19 response some countries have used AI and Big-Data to support provision of temporary support. However, countries are still struggling in setting boundaries for the utilization of AI. Just last October (2023), the US issued an executive order to regulate AI, and last June the European Parliament adopted its negotiating position on the AI Act to make to ensure that AI systems used in the EU are safe, transparent, traceable, non-discriminatory, and environmentally friendly. In both cases, it is recognized that AI systems can be applied in various contexts, but they pose risks to users, and different levels of regulation are warranted based on varying risk levels. 95. However, in addition to regulations governing access to private data by government entities, it is crucial to note that aligning the social protection unit of assistance, which can be an individual, household, or family, with big data is not always straightforward. And errors can be significant, particularly when the unit of assistance is either the household or the family, as most of the available Big Data, such as satellite images and CDR, are not accessible for such assistance units. Most examples utilizing Big Data combined with AI/machine 31 learning algorithms are designed for reaching individuals, as seen recently in the COVID-19 response (see Blumenstock, 2020 and Okamura et al., 2024), or for small areas/grid levels (relevant for supporting geographical targeting), without the capability to aggregate information for a household. In those examples, program administrators could not ensure that only one household member was receiving financial support or guarantee the inclusion of people without access to cellphones, leading to digital exclusion (see below). Moreover, cell-phone users may have multiple phones or share them among individuals or even households. Hence, given the significance of the data from call records associated with the SIM card or phone number the issue becomes bigger and more complex than usually assumed. 39 96. Systematic Bias is a general concern (see digital exclusion sessions below) but more important than ever in the context of the increasing use of AI and Big Data. AI (though machine learning) and Big Data are already being combined with traditional data to improve poverty maps and help predict which households and areas are more at risk of natural disasters. In identifying populations to be treated in a crisis or data-scarce environment, such as post-conflict contexts, using big data for determining eligibility may be one of the only options and an appropriate one. Additionally, these technologies are used to enhance the precision of proxy-means methods accuracy, as observed in Costa Rica and Colombia.40 But, one must pay attention to avoid bias in prediction. 97. Machine learning models data (such as regular survey data and Big Data) are needed to algorithms to learn and predict who should be selected to participate in programs. When the data they train on does not represent the whole population, model predictions can be biased. Machine learning algorithms also risk perpetuating inequalities and bias (exclusion) against certain groups as modeling may be based on data that reflects historical biases as some groups or populations are not well represented, or even misrepresented, in the machine learning used data. For example, Grosh et al. (2022) discussed how machine learning algorithms could discern that variables such an ethnicity are good predictors of outcomes. But except for certain affirmative action programs, in most places, ethnicity-based 39 For example, in Togo, the Novissi program used both AI and Big-Data to support response to the COVID-19 pandemic, utilizing call-detail records (CDR), phone data collection, harnessed artificial intelligence and machine learning to prioritize poorest areas and individuals to receive support. In the Democratic Republic of Congo, the Step-Kin used satellite imagery with a 100x100 meter grid to identify hotspots, and online application that selected individuals from 1 out the four cell-phone companies (limited due to procurement issues) after filtering out subscribers with higher incomes: those owning a smart phone, having a prepaid data plan, monthly expenditures of over $5 per month or with international calls. 40 Grosh et al. (2022) 32 targeting would not be acceptable. Or the algorithm could pick up other aspects of historical discrimination, related for example, to ethnic ghettos or redlining. As computers cannot distinguish between ethical and unethical decisions yet, data scientists need to be aware of historical biases and consider how they can be addressed. Therefore, careful checks need to be put in place to ensure eligibility assessments do not disadvantage sub-populations. The marginalized groups of interest may often be exactly the ones missing from the data sources used, primarily from Big Data.41 98. Big Data and AI may raise significant human rights issues, although work is underway to try and provide privacy guarantees while still allowing for public good applications. 42 In the case of Togo, the academic team carefully tested for demographic parity and systematic prediction errors for many different subgroups such as gender, ethnicity, religion, oversight, and audits. But in a System, such type of testing should be done in house by the System administrators to avoid potential privacy concerns that can be raised when private vendors deal with such information. 99. Without careful management, AI algorithms can encode bias, hence skilled human resources must be involved in the System design. As well, there are several concerns and guidelines raised by UN agencies. On Identity management, including those of WFP with regards to data sharing - "We do not share people’s personal data if harm could come to them-; and privacy - "We respect people’s privacy by collecting only the minimum data required to ensure that they receive assistance, by protecting that data as best we can, and by deleting it when it isn’t needed anymore". 43The debate between AI opportunities and risks has taken place and continues to evolve. On the one hand, people are worried about the observed risks of AI on humanity, and advise that we need to slow things down, whereas others believe that AI will solve several problems the world faces, and we could support its advancement. There is room in the middle. H. Digital Exclusion 100. The reliance on technology can increase systematic bias against certain population groups in the absence of a human-centered design approach. When technology is considered 41 GIZ (2020) careful analysis of using big data in social protection shows that despite having multiple usages, data protection, including which data the state can legitimately use to determine eligibility, is just one of the many questions for this promising but still immature field. 42 Tables 6-10 in Grosh et al. (2022) 43 UN 2023 Department of economic and social affairs Guidelines on the Legislative Framework for Civil Registration, Vital Statistics and Identity Management Systems. 33 a replacement rather than a complementary tool to humans—the core actors that can help populations overcome issues of information, agency, or digital divide in the provision of social services—systemic biases may be exacerbated. Moreover, one must account for the fact that many of the desired population may still be “invisible” in the eyes of other administrative data systems. 101. Being invisible to is also caused by poor, non-reliable and not affordable connectivity in both rural and remote areas where many of System users reside, leading to systemic exclusion and digital exclusion. In developing countries, internet access remains very expensive and slow, coverage of individuals with access to internet is low and leads to exclusion. 44 For example, in Bhutan, its bandwidth cost is currently almost double that of India and Bangladesh, and access to fixed broadband remains below 1 percent of households and unaffordable for smaller businesses and the poor. In the Democratic Republic of Congo, only one cell phone company, the Africell, which was also a newcomer in the market, was proactive with putting in extra effort and resources to reach STEP-KIN beneficiaries as post- COVID emergency program but its small coverage (16 percent of market) had impacted the ability of administrators to reach out many of the affected ones. 102. The combination of higher prices and low demand, combined with power outages, are obstacles to provide better access of internet by telecoms. In some countries even if they have the infrastructure in place or access to purchase additional fiber optic cables and other infrastructure, globalization and weak currency blocks telecoms and governments to invest on network to make internet more affordable and then more attractive to poorest and vulnerable ones. Therefore, it is important as part of the System development to discuss with other agencies involved in providing wide digital infrastructure across the country as part of the institutional arrangements and to properly access the likely negative impacts of such poor infrastructure to the System under development. As more users of System would be in remote and rural areas, economies of scale can be generated with proper institutional coordination and governance. I. Systemic Exclusion 103. As technology promotes the use of devices for information collection, electronic applications, interoperability through unique "keys," electronic payments, etc., individuals 44 Sub-Saharan Africa suffers from overpriced internet at Le Monde 2023. https://www.lemonde.fr/en/pixels/article/2023/05/05/sub-saharan-africa-suffers-from-overpriced- internet_6025513_13.html 34 without access to basic technology, electricity, documents, or digital literacy can face exclusion by design. Designing the system as an enabler to improve efficiency, rather than a blocker for those without access or knowledge of how to use technology, is essential. Digital accessibility does not replace physical accessibility. Internet communication (including though social platforms) does not replace radio, TV or community sessions/camps. Direct online applications using cellphones apps or websites does not replace face-to-face declaration of information in registration sites. Moreover, many countries still have limited coverage of National IDs, hence requirements of presenting documentation for the sake of interoperability of systems can exclude or delay access for many, while many of socially excluded or marginalized groups are also invisible in regard to existent government systems, as they are not part of any service offered or system run by the government and hence excluded as per the discussion on the AI and Big Data above on systematic bias. By having a thoughtful combination of “digital” and no-digital approaches for communications, outreach, and intake and registration, while keeping required documentation to a minimum remains crucial to reach out to those in need, mostly the marginalized and in remote areas, where technology penetration is still limited. In addition, requiring applicants to declare and prove their identity depending on the country’s identity ecosystem and its degree of integration with the programs. Of course, the drive toward higher coverage of foundational ID systems hugely facilitates some goals and processes of social protection. But across the board it is important to offer access to registration through the provision of alternatives for those who do not have the first-choice document. 45 Court cases in several countries have upheld the right to services without the requirement of ID, notably India which is not alone in this regard. 46 J. Digital literacy/exclusion (access) 104. Many in society are still not benefitting (to the potential) from the digital revolution and the use of digital technologies (e.g., internet as many can have access to it but still limited skills to be able to navigate and benefit from). Digital access and digital skills are still affecting people and thus create a digital gap between those connected and those not connected. In developing countries libraries are often a place for the most disadvantaged groups. For 45 US identity is proved through the social security number and driving licenses. In Brazil application in the cadastro unico requires one document with a picture that can be the state level ID or Labor Card, and another document such as social security number or voting number. 46 https://www.apc.org/en/news/extreme-poverty-and-digital-welfare-new-report-un-special-rapporteur- extreme-poverty-raises 35 example, during COVID, in US, many were sitting in parking lots to be connected; 47 schools and districts helped students meet these needs by providing computers or paying for home internet access; 48 and studies as that of Ogundari (2023) show that the observed effects of access to technology on learning hours highlighted the race and ethnicity disparities in American society’s digital divide linked to access to technology. Along similar lines, ONS (2019) indicates that older people are the largest proportion without internet users, and by age group, as the age bracket increases more likely to have lower digital skills. Some parts of the internet are also unfriendly for people with disabilities. While there have been improvements in access to the internet for people with disabilities, Systems must be designed to facilitate navigation, such as incorporating features for visually impaired people. Even in physical centers, the centers must be adapted to accommodate individuals with disabilities in person. 105. As more people have access to technology one can assume that its accessibility is universal and can easily improve efficiency of programs, but digital divide, digital illiteracy, coverage, physical disabilities and equitable access to technology remains unbalanced in and amongst developing countries. These countries are at various stages of technological development, exhibiting disparities in technology use, accessibility and coverage. Unequal distribution and penetration of technology and connectivity within each country need to be considered, and failure to account for these variations may lead to increased inequality and inefficiencies, even as technology advances in certain regions. This “penetration” of technology leads to a digital divide that can generate significant exclusion and inefficiencies during implementation as many may have no (or limited) access to technology, and some can also be “digitally or technologically illiterate.” Physical disabilities are to be considered while designing the MIS as some people would need special tools to be able to benefit from the available technology. For example, online applications should be designed with consideration for assisted technology to accommodate blind and low-vision applicants. Single application window offices should be equipped to serve all types of disabilities, including the provision of access ramps and special bathrooms. 106. Poor broadband and mobile infrastructure should be taken into consideration, as social protection applications must operate in such environments. Therefore, they cannot be resource-intensive or require premium gadgets. Finally, as connectivity/penetration of 47 https://www.nytimes.com/2020/05/05/technology/parking-lots-wifi-coronavirus.html https://nces.ed.gov/blogs/nces/post/students-internet-access-before-and-during-the-coronavirus-pandemic- 48 by-household-socioeconomic-status 36 technology in remote areas remains an issue in most developing countries, one must have alternative design and implementation processes for ensuring that all populations benefit from investments in technology. For example, in Bhutan, as the Accelerated Maternal and Child Health program is being developed, its IT system relies on existing internet penetration to deploy the online system, while in remote areas, an offline system that can upload information to the center at a given interval using remote access tools is being made available. V. The importance of Human resources 107. Systems are fundamentally about human interactions. While technology can enhance efficiency and streamline processes, it can never fully replace the need for human interactions at different levels. 108. For some people, there is a risk of misunderstanding technology as a replacement for essential human resources 49, especially in the context of AI systems. While AI systems will probably have the potential at some point to conduct certain tasks, the same may not be assumed about jobs. For any program to be successful, it requires having skilled and equipped personnel to be at the forefront of program implementation regardless of the level of technology in place. 109. Maintaining human interactions at registration point allows those who lack access or skills to use the digital platforms. Lindert et al. (2020) highlights that human resources are a fundamental requirement to good program implementation as they must have the knowledge and skills needed to do their job, resolving problems and handling complaints, interacting with families, village heads, service providers, and local officials as the “face of the program,” and so on. Therefore, skilled and equipped professionals must be in place to deal with the day-to-day activities and program implementation effectively. 110. Physical offices, public access kiosks, and mobile outreach services are essential for those who lack effective access to digital services. This can be due to various reasons, including limited internet coverage, poor service quality, affordability issues, lack of personal devices, limited digital literacy, and language barriers. These traditional service channels are crucial to ensure inclusivity and reach a broader segment of the population. As part of the registration process, human interaction is essential for engaging people and ensuring access 49 Grosh et al. (2022) 37 to high-quality, accurate information. Proper training is crucial for data collection, covering both the methods of gathering information and proficiency in using the software. This ensures that administrators are well-equipped to navigate through the system efficiently. 111. Most importantly, AI and Big Data, and technology in general do not replace the need to invest in people and human process as Joshua Blumenstock (CEGA Faculty Co-Director) 50 highlighted in one of its interviews on AI and Big Data 51 when he said: “The algorithms are sort of the shiny object, and they receive a lot of attention. But when it comes to actually implementing social protections, going the last mile to put money in the hands of people who need it, the algorithms are just one small link in a much larger chain of humanitarian assistance. Most of the other links are human. Algorithms can help surface relevant information, but humans must decide what to do with it.” 112. Moreover, massive data manipulation and AI use requires System administrators to have skilled human resources for developing, training and auditing models when developing AI tools for improving accuracy and data security while reducing bias. Therefore, use of technology requires skilled labor force working in the government to avoid lock-ins to private sector developers, which are costly and raises data privacy concerns. 113. This is the case for functional programs around the world that massively invest in technology and human resources simultaneously. For example, in Brazil, despite all the interoperability available, Social Assistance Reference Centers (CRAS) assist people to access social programs at the municipal level. Each CRAS is equipped with the necessary human resources and infrastructure to establish local partnerships and mobilize communities. They play a vital role in informing the community about the importance of investing in human capital for increased productivity and accessibility to various services, programs, projects, and social assistance benefits. As trained staff at CRAS enter initial contact details into the System during face-to-face sessions, where individuals provide self-declarations of their socio-economic conditions in the Cadastro Único, people are educated about verification measures, and false declarations may result in the suspension of benefits and penalties, affecting all family members' eligibility for assistance. 114. Building an effective System requires a diverse skill set, and key to its success are the human resources responsible for both operating the system and serving as its public face in 50 Center for Effective Global Action at University of California - Berkeley (https://cega.berkeley.edu/) 51 https://news.berkeley.edu/2020/06/02/satellite-images-phone-data-help-guide-pandemic-aid-in-at-risk- developing-countries 38 interactions with individuals. Logisticians play a crucial role in planning efficient delivery systems, while social workers or community development officers act as intermediaries, forging connections between ministries, communities, and individuals, placing them at the heart of a well-functioning System. Moreover, using technology as a change tool rather than a solution to a problem has shown better success in digital projects. Generally, the tenants of change management include people, processes, and tools (such as technology). Striking the right balance between all three will generally ensure higher success factors on transformation projects. The transformational potential of technology will be solidified, and one would be able to harness value out of implementations through successfully transitioning along the change path with these principles of change, especially when the transformation is aligned to programmatic objectives such as better targeting, less exclusion errors, traceability, transparency, and feedback loops. VI. Conclusion 115. Increasing access and use of digital services can improve the efficiency of systems and access to benefit and services. However, technology is not a panacea and not risk free. Technology can increase and improve the quality of the interactions between people and institutions, and in the end improve social equality and equal access. And as countries increase investments in systems and use of technology, this raises red flags and risks to be considered and addressed. Technology helps interoperability and integration of administrative systems, improving data quality and individual experience, such as outreach and monitoring of social protection programs. The expansion of the coverage of unique IDs where more people hold and use a foundational or functional ID, especially e-IDs, and increased computing power is making it far easier to create integrated or interoperable data systems that lower costs and increase the dynamism of systems. As data coverage increases and quality improves, more countries will meet the minimum conditions to move toward dynamic operable systems that also benefit from big data and artificial intelligence advances to accurately reach and provide access to benefits and services. Moreover, investment in technology must be accompanied by investment in human resources. 116. There is not a single recipe to guide the use of technology, but some challenges (and risks) are well-known. Main challenges to overcome include absence of political will, limited legal framework and weak organization/coordination. Main technical challenges relate to having data systems specificities well planned in advance; defining protocols for safe data 39 collection and secure use of data, including validation; ensuring consent for disclosure of information; protection against data breaches; development of safeguards, security protocols including firewalls, backups and recovery to protect data; increase use of big data and AI which can lead to systemic bias; digital exclusion and digital literacy that affects most disadvantage groups in general; and limited invest in human resources and loss of human aspect of social programs. Therefore, contextual customization matters when designing and building systems and using technology. 117. There are several principles that apply, but in the end what makes sense in one country with regards to using more technology may not necessarily apply to another as one must consider specific, unique features of the setting, i.e., the coverage of national and digital IDs, platforms and infrastructure and penetration of technology, the characteristics of the population with and without access to technology, as well as the political economy of boosting efficiency of the System. 118. In conclusion, technology can improve delivery systems and work as a bridge between central administration and potential population. Nevertheless, while building these bridges human resources and local authorities play an important role as they are the face of the programs and must be properly trained and equipped to support the system. Participation of government staff (whether locally posted federal agencies or at the municipal level) that are equipped with appropriate resources and incentives and proper citizen engagement legitimizes the process and help improve program outcomes. That said, the development of solid systems takes a diverse skill set, including investing in skilled IT engineers and developers for planning robust delivery systems and social workers or community development officers that can facilitate connections between ministries, communities, and individuals. Moreover, advocates, organizers, technocrats, and elected officials play pivotal roles in fostering consensus for robust social protection policies. The integration of technology further augments service delivery, leading to enhanced efficiency and effectiveness. 40 Bibliography Aiken, Emily; Ohlenburg, Tim. 2023. “Novel Digital Data Sources for Social Protection: Opportunities and Challenges.” www.socialprotection.org/sites/default/files/publications_files/20230327_novel_data_sou rce_for_SP_0.pdf Alston, Philip. 2019. “Report of the Special rapporteur to the UN on extreme poverty and human rights.” A/74/48037. Bashir, Sajitha; Dahlman, Carl J.; Kanehira, Naoto; Tilmes, Klaus. 2021. “The Converging Technology Revolution and Human Capital: Potential and Implications for South Asia.” South Asia Development Forum; Washington, DC: World Bank Blumenstock, Joshua. 2020. “Machine learning can help get COVID-19 aid to those who need it most.” Nature. 2020 May 14. doi: 10.1038/d41586-020-01393-7. Bowen, Thomas Vaughan; Del Ninno, Carlo; Andrews, Colin; Coll-Black, Sarah; Gentilini, Ugo; Johnson, Kelly; Kawasoe, Yasuhiro; Kryeziu, Adea; Maher, Barry Patrick; Williams, Asha M. 2020. “Adaptive Social Protection: Building Resilience to Shocks.” International Development in Focus Washington, D.C.: World Bank Group. http://documents.worldbank.org/curated/en/579641590038388922/Adaptive-Social- Protection-Building-Resilience-to-Shocks Delgado Francisco; Oton, Salvator; Julian Ruggia, Raul; Hilera, José R.; Gutiérrez, José M. 2013. “Proposal of an Interoperability Model for Social Security Information Systems.” DOI: 10.5220/0004438204510458 In Proceedings ofthe15th International Conference on Enterprise Information Systems (ICEIS-2013), pages451-458 EIU. 2024. https://www.eiu.com/industry/telecommunications/subsector/next-generation- technologies GIZ. 2019. “Building an integrated and digital social protection information system.” https://www.giz.de/de/downloads/giz2019-en-integrated-digital-social-protection- information-system.pdf. Deutsche Gesellschaft für Internationale Zusammenarbeit (GIZ) GIZ. 2020. “AI in Social Protection – Exploring Opportunities and Mitigating Risks.” Deutsche Gesellschaft für Internationale Zusammenarbeit (GIZ) 41 https://socialprotection.org/sites/default/files/publications_files/GIZ_ADB_AI%20in%20soc ial%20protection.pdf GIZ .2023a. “Novel digital data sources for social protection: opportunities and challenges.” https://socialprotection.org/sites/default/files/publications_files/20230327_novel_data_so urce_for_SP_0.pdf. Deutsche Gesellschaft für Internationale Zusammenarbeit (GIZ) GIZ .2023b. “Pushing the Boundaries of Social Protection Administrative Data Analysis and Use.” https://socialprotection.org/sites/default/files/publications_files/GIZ- %20Use%20Admin%20Data%20Social%20Protection.pdf Grosh, Margaret; Leite, Phillippe; Wai-Poi, Matthew; Tesliuc, Emil. 2022. “Revisiting Targeting in Social Assistance: A New Look at Old Dilemmas.” Human Development Perspectives” © Washington, DC: World Bank. http://hdl.handle.net/10986/37228 Ijjasz-Vasquez, Ede; Jongman, Brenden; and Suarez, Pablo. 2017. “From algorithms to virtual reality, innovations help reduce disaster risks and climate impacts.” https://blogs.worldbank.org/en/sustainablecities/algorithms-virtual-reality-innovations- help-reduce-disaster-risks-and-climate-impacts Karippacheril, Tina George; Alberro Encinas, Luis Iñaki; Cardenas Martinez, Ana Lucia; Daly, Conrad; Suri, Satyajit. 2024. “Playbook on Digital Social Protection Delivery Systems: Towards Dynamic Inclusion and Interoperability.” ©Washington, DC: World Bank. http://hdl.handle.net/10986/41527 Leite, Phillippe; George, Tina; Sun, Changqing; Jones, Theresa; Lindert, Kathy. 2017. “Social Registries for Social Assistance and Beyond: A Guidance Note & Assessment Tool.” © Washington, DC: World Bank. Social Protection and Jobs, Discussion paper No. 1704. July 2017. http://documents.worldbank.org/curated/en/698441502095248081/Social- registries-for-social-assistance-and-beyond-a-guidance-note-and-assessment-tool Leite, Phillippe; and Karippacheril, Tina George. 2019. “Safety Nets Core Course presentations.” https://www.worldbank.org/en/events/2019/03/18/social-protection-and- jobs-core-courses-2019 42 Lindert, Kathy; Karippacheril, Tina George; Rodriguez Caillava, Inés; Nishikawa Chavez, Kenichi. 2020. “Sourcebook on the Foundations of Social Protection Delivery Systems.” © Washington, DC: World Bank. http://hdl.handle.net/10986/34044 Lowe, Christina; Rigolini, Jamele; Solbes Castro, Lucia; Bastagli, Francesca. 2023. “Pathways toward digitalization in Social Protection and Labor (SPL) service delivery.” © Washington, DC: World Bank. Social Protection and Jobs, Discussion paper No. 2307. August 2023. http://documents.worldbank.org/curated/en/099100523151038686/P173530036a7f30f60 899c06a6172040bf3 Mukherjee, Anit Nath; Rojas, Bermeo; Ximena, Laura; Okamura, Yuko; Muhindo, Jimmy Vulembera; and Bance, Paul G. A. 2023. “Digital-first Approach to Emergency Cash Transfers: STEP-KIN in the Democratic Republic of Congo.” Social Protection and Jobs Discussion Paper; no. 2302 Washington, D.C.: World Bank Group. OECD. 2019. “Enhancing Access to and Sharing of Data: Reconciling Risks and Benefits for Data Re-use across Societies.” OECD Publishing, Paris, https://doi.org/10.1787/276aaca8-en Ogundari K. 2023. “Student access to technology at home and learning hours during COVID- 19 in the U.S.” Educ Res Policy Prac. 2023 May 12:1–18. doi: 10.1007/s10671-023-09342-7. Epub ahead of print. PMCID: PMC10176282. Okamura,Yuko; Ohlenburg,Tim Julian; Tesliuc,Emil Daniel. 2024. “Scaling up Social Assistance Where Data is Scarce: Opportunities and Limits of Novel Data and AI.” Social Protection and Jobs Discussion Paper; No. 2402 Washington, D.C. : World Bank Group. http://documents.worldbank.org/curated/en/099050724145524418/P17191311f545a0971 b3db17f9d6820d240 Office for National Statistics. 2019. “Internet access – households and individuals, Great Britain: 2019.” https://www.ons.gov.uk/peoplepopulationandcommunity/householdcharacteristics/homei nternetandsocialmediausage/bulletins/internetaccesshouseholdsandindividuals/2019) Red Cross. 2017. “Forecast-based financing: case studies from Togo and Uganda.” https://www.climatecentre.org/publications/3161/forecast-based-financing-case-studies- from-togo-and-uganda/ 43 Sepulveda Carmona, Magdalena. 2018. “Is biometric technology in social protection programmes illegal or arbitrary? An analysis of privacy and data protection.” Extension of Social Security Working Paper 59, International Labor Organization, Geneva. https://www.researchgate.net/publication/325909014_Is_biometric_technology_in_social _protection_programmes_illegal_or_arbitrary_An_analysis_of_privacy_and_data_protecti on Silva, Veronica; Heidi Berner, Luis Díaz Silva, and Javiera Troncoso. 2018. “Registro Social de Hogares de Chile.” http://www.desarrollosocialyfamilia.gob.cl/storage/docs/RSH_paper_2.pdf. Sluchynskyy, Oleksiy. 2019. “Better Data for Better Social Programs and Policies Enabled by e-Government Solutions: Case Studies.” World Bank, Washington, DC. https://thedocs.worldbank.org/en/doc/674741548878149565- 0160022019/original/10.45amJan15DeliverySystemSluchynskyInteroperabilityanddatashari ng.pdf SPIAC-B. 2022. “Implementation Guide – Good Practices for Ensuring Data Protection and Privacy in Social Protection Systems: A guide for practitioners working and advising in low- and middle-income countries.” https://socialprotection.org/discover/publications/implementation-guide-%E2%80%93- good-practices-ensuring-data-protection-and-privacy Turkey’s Integrated Social Assistance System Report, https://documents1.worldbank.org/curated/pt/515231530005107572/pdf/Turkey-SA- summary.pdf UNICEF. 2019. “Programme Guidance: Strengthening Shock Responsive Social Protection Systems” https://www.unicef.org/media/68201/file/Practical-Guidance-to-Support-Work- on-Shock-Responsive-Social-Protection.pdf UNICEF. 2021. “Better integration of social protection and humanitarian information systems for shock response.” https://www.unicef.org/media/100016/file/Technical%20note%20- %20Information%20Systems.pdf 44 UN Personal Data Protection and Privacy Principles. https://archives.un.org/sites/archives.un.org/files/_un-principles-on-personal-data- protection-privacy-hlcm-2018.pdf United Nations Systems. 2022. “Principles for the Ethical Use of Artificial Intelligence in the United Nations System.” https://unsceb.org/sites/default/files/2022- 09/Principles%20for%20the%20Ethical%20Use%20of%20AI%20in%20the%20UN%20Syste m_1.pdf WFP. 2021. “World Food Programme Strategy for Support to Social Protection – 2021” https://www.wfp.org/publications/world-food-programme-strategy-support-social- protection-2021 WFP. 2022. “How blockchain can power efforts to empower women and girls in Bangladesh.” World Food Programme (wfp.org) https://www.wfp.org/stories/how- blockchain-can-power-efforts-empower-women-and-girls- bangladesh#:~:text=%E2%80%9CTechnologies%20such%20as%20blockchain%20can,WFP's %20Country%20Director%20for%20Bangladesh. World Bank. 2021. “Enrollment and Eligibility Process of Brazil’s Auxílio Emergencial: Data Processing and Use of Administrative Registries.” © Washington, DC. http://hdl.handle.net/10986/37256 License: CC BY 3.0 IGO. https://hdl.handle.net/10986/37256 World Bank. 2022. “Charting a Course Towards Universal Social Protection: Resilience, Equity, and Opportunity for All.” © World Bank Group, Washington, DC. https://hdl.handle.net/10986/38031 World Bank. 2023. “Responsive by Design: Building Adaptive Social Protection Systems in South Asia.” Washington, D.C.: World Bank Group. http://documents.worldbank.org/curated/en/099192003212311865/P17869108018a10c50 822c05d3eea25a5 45 Social Protection & Jobs Discussion Paper Series Titles 2025 No. Title 2505 A slippery slope: the opportunities and risks of digital approaches and technology in Social Protection Systems 2504 De Jure and De Facto Coverage of Parental Benefits in Nepal 2503 Awareness, Access, and Perceptions around Parental benefits among Urban Argentinians 2502 Regulating Markets So More People Find Better Jobs 2501 São Tomé and Príncipe Unpacking Migration Dynamics: Critical Issues and Policy Recommendations To view Social Protection & Jobs Discussion Papers published prior to 2021, please visit www.worldbank.org/sp. ABSTRACT Advances in technology have the potential to enhance social protection services delivery but come with risks, such as data privacy concerns, exclusion, and biases. To use digital technologies effectively, strong frameworks, infrastructure, and capacity are essential. Without these, technology may inadvertently harm rather than benefit the intended populations. Technology can improve access, outreach, training, monitoring, and secure payments, among others, but risks must be managed by clearly defining roles and responsibilities among stakeholders. Hence, policymakers must understand challenges and define processes before selecting technology, viewing it as a tool to complement, not replace, non-digital services. Therefore, technology should support human resources in social protection programs, and aligning innovation with effective safeguards can maximize its potential for equitable and responsible outcomes. JEL CLASSIFICATION D60, D70, D80, D81, D83, I38, O33, O35, O38, and O57 KEYWORDS Artificial Intelligence, Big Data, machine learning, social registries (aka unified registries, single registries, unique registries, registration and eligibility systems), social protection delivery system, social protection system, delivery systems, delivery chain, social assistance, services, transfers, social protection, data security, data privacy, data collection. ABOUT THIS SERIES Social Protection & Jobs Discussion Papers are published to communicate the results of The World Bank’s work to the development community with the least possible delay. This paper therefore has not been prepared in accordance with the procedures appropriate for formally edited texts. For more information, please visit us online at www.worldbank.org/socialprotection