Decentralizing AI Economics for Poverty Alleviation: Web3 Social Innovation Systems in the Global South

Abstract

Artificial Intelligence (AI) is increasingly framed as a driver of economic transformation, yet its capacity to alleviate poverty in the Global South remains contested. This article introduces the notion of AI Economics—the political economy of value creation, extraction, and redistribution in AI systems—to interrogate h ow innovation agendas intersect with structural inequalities. This article examines how Social Innovation (SI) systems, when coupled with decentralized Web3 technologies such as blockchain, Decentralized Autonomous Organizations (DAOs), and data cooperatives, may challenge data monopolies, redistribute economic gains, and support inclusive development. Drawing on Action Research (AR) conducted during the AI4SI International Summer School in Donostia-San Sebastián, this article compares two contrasting ecosystems: (i) the Established AI4SI Ecosystem, marked by centralized governance and uneven benefits, and (ii) the Decentralized Web3 Emerging Ecosystem, which promotes community-driven innovation, data sovereignty, and alternative economic models. Findings underscore AI’s dual economic role: while it can expand digital justice, service provision, and empowerment, it also risks reinforcing dependency and inequality where infrastructures and governance remain weak. This article concludes that embedding AI Economics in context-sensitive, decentralized social innovation systems—aligned with ethical governance and the SDGs—is essential for realizing AI’s promise of poverty alleviation in the Global South.

1. Introduction

Artificial Intelligence (AI) has rapidly become a transformative force in global development, revolutionizing industries and urban infrastructures [1]. Often viewed as a driver of economic growth, the technology prompts a critical research question: Can the emerging field of AI Economics—the study of how AI redistributes value, labor, and resources—contribute to poverty alleviation in the Global South? This question is central to debates on technology’s role in addressing entrenched socio-economic inequalities and the structural challenges of deploying AI in regions with fragile infrastructures and socio-political conditions [2,3,4].

Within this context, AI Economics can be understood as the political economy of algorithmic systems—examining how machine-learning models reconfigure value creation, labor, resource allocation, and governance structures. Rather than focusing solely on digitalization or platform logics, AI Economics foregrounds how predictive and optimization systems reshape institutional capacities, market coordination, and distributional outcomes. This perspective is particularly relevant for the Global South, where debates on AI increasingly intersect with questions of digital sovereignty, infrastructural vulnerability, and asymmetries in computational power. Understanding whether AI-enabled systems can meaningfully contribute to poverty alleviation therefore requires not only technological assessment but also an examination of governance models, value flows, and the conditions under which communities might retain agency over the economic and social benefits generated by AI.

In the Global North, AI optimizes public services, decision-making, and innovation [5], yet its potential for equitable impact in the Global South—where disparities in digital resources and economic power are stark—remains uncertain [6]. This divide underscores the stakes of AI Economics for the Global South, where disparities in digital resources and power risk exacerbating inequalities rather than reducing them [7,8]. AI’s dual role—as an enabler of progress and a disruptor of traditional systems—underscores its promise and its risks for the Global South [9]. AI could address critical challenges in marginalized communities, improving healthcare, agriculture, and education [10,11,12]. However, reliance on AI also risks exacerbating the digital divide, with control concentrated in the Global North [13].

Countries in the Global South, particularly low- and middle-income regions, face barriers that limit the potential of AI Economics to deliver inclusive growth. These include the absence of robust infrastructures, gaps in digital literacy, and insufficient governance frameworks [14,15,16]. In this context, the economic dimension of AI is inseparable from broader debates on digital inclusion and equity [17,18]. Without intentional design, AI may reinforce “digital colonialism,” where value extraction flows to the Global North while costs and risks remain local [19,20]. Therefore, discussions of AI in the Global South must consider broader debates on digital inclusion and equity [21,22,23].

AI’s potential lies in its ability to deliver targeted solutions to local needs [24,25]. In agriculture, AI can optimize crop yields, monitor soil, and predict weather, enhancing food security and sustainability (i.e., https://app.agrimetrics.co.uk/ (accessed on 1 November 2025)). In healthcare, it can improve diagnostics, streamline services, and expand access to care in resource-limited regions (i.e., www.dimagi.com). However, such applications face hurdles, including inadequate localized data, insufficient digital infrastructure, and algorithmic biases [26,27,28,29]. AI is sometimes mischaracterized as a “magic tool” that effortlessly solves societal challenges, a perception driven by media and tech industry narratives [30,31]. Yet, for AI to meet its promises in the Global South, foundational investments in infrastructure, education, and governance are essential [32]. AI cannot operate effectively without supportive socio-technical ecosystems.

Social Innovation (SI) frameworks, such as AI for Social Innovation (AI4SI (The abbreviation AI4SI used by the World Economic Forum (WEF) can stand for both AI for Social Innovation and AI for Social Impact, depending on the context. While these terms are closely related, they emphasize slightly different aspects: (1) AI for Social Innovation focuses on how AI can be applied to create new solutions, strategies, or models that address social challenges, particularly through creative and disruptive approaches. It emphasizes the process of developing innovative, AI-driven solutions to improve societal well-being, equity, and justice. (2) AI for Social Impact emphasizes the outcomes or effects of AI on society, particularly how AI-driven technologies can lead to measurable positive changes in areas like healthcare, education, sustainability, or economic empowerment. The term is more focused on the results of AI applications and their direct impact on communities and individuals. In practice, WEF uses AI4SI to refer to initiatives that address both the innovative applications of AI and their societal impact. Therefore, AI for Social Innovation and AI for Social Impact can be seen as interchangeable under the AI4SI framework, with both innovation and impact being integral to the initiative’s goals)), aim to address these challenges by creating AI solutions aligned with local needs and capacities [33,34,35,36,37]. Through AI-driven SI, marginalized communities may gain enhanced access to public services and more participatory governance models [13,38,39]. Fieldwork initiatives, like the AI4SI International Summer School in Donostia-San Sebastián https://www.uik.eus/en/activity/artificial-intelligence-social-innovation-ai4si (accessed on 1 November 2025); [40], explore how decentralized Web3 technologies (e.g., blockchain, DAOs, data cooperatives) can promote equitable governance and data sovereignty in the Global South [41,42]. Such efforts reveal AI’s potential to challenge centralized power and open pathways to inclusive development [43].

A key insight is the importance of adapting AI to the socio-political realities of the Global South [44]. The success of AI-driven solutions depends on the readiness of the local environment [45,46,47,48]. Without robust infrastructure and governance mechanisms, AI risks perpetuating “digital colonialism,” where data control is held by Global North corporations [19,20]. This underscores the need for a decolonial approach that respects local knowledge and community needs [49]. Ethical concerns, including data sovereignty, privacy, and algorithmic bias, are central to deploying AI as a tool for poverty alleviation [50,51,52,53]. In regions with less resilient governance, AI’s deployment requires careful ethical consideration [54].

The question of AI’s potential for poverty alleviation in the Global South goes beyond technology alone [55,56,57]. A holistic approach combining SI, equitable governance, and sustained local investment is essential [58]. AI must be viewed as an empowering tool, anchored in principles of digital justice [59]. This article argues that AI’s potential for inclusive development lies in frameworks ensuring equitable access and community engagement [60].

This article therefore situates AI not simply as a technological tool but as an evolving economic system with implications for justice, sovereignty, and sustainable development. By analyzing two contrasting ecosystems—the Established AI4SI Ecosystem, dominated by centralized governance and economic concentration, and the Decentralized Web3 Emerging Ecosystem, centered on community-driven innovation—the study explores how decentralizing AI Economics can create new pathways for poverty alleviation in the Global South.

Consequently, this article makes four main contributions:

(i)

Conceptualization of AI Economics for poverty alleviation: This article clarifies AI Economics as a political economy framework for understanding value creation, extraction, and redistribution across AI systems, with attention to distributional outcomes in the Global South.

(ii)

Comparative analysis of two innovation ecosystems: This article offers a systematic comparison between the Established AI4SI Ecosystem and the Decentralized Web3 Emerging Ecosystem, highlighting their contrasting governance structures, value flows, and social innovation logics.

(iii)

Identification of structural enablers and constraints: This article examines how centralized legacy infrastructures and emerging decentralized architectures each generate distinct opportunities and risks for inclusive development, focusing on data sovereignty, institutional capacity, and participatory governance.

(iv)

Proposal of a hybrid, ethically governed model: This article outlines a hybrid configuration that integrates the stability and regulatory grounding of centralized systems with the participatory and redistributive potential of Web3 tools such as DAOs and data cooperatives, offering a pathway for future policy experimentation and research.

While this paper refers to the Global South as a broad geopolitical category, it is important to acknowledge that it encompasses regions with highly diverse infrastructural capacities, governance systems, regulatory environments, economic structures, and cultural contexts. Countries in Africa, Latin America, and Asia experience AI deployment under markedly different conditions—ranging from bandwidth constraints and uneven institutional capacity to variations in state–citizen relations, trust in digital systems, and levels of digital literacy. As a result, the pathways through which AI Economics might support poverty alleviation differ substantially across regions, and no single model can be assumed to apply uniformly. The case studies included in this paper should therefore be interpreted as illustrative examples situated within their respective contexts rather than as representative of the Global South as a whole. The paper therefore focuses on the institutional and governance dynamics that create potential pathways towards poverty alleviation, rather than on measuring direct poverty outcomes.

The article is structured as follows. After this introduction, which presents the research question, the article presents the state-of-the-art research and policy framework of AI4SI to unpack the research question: (i) a literature review highlighting key literature, (ii) a policy analysis, and (iii) several findings and insights stemming from the action research process carried out through the AI4SI International Summer School in Donostia-San Sebastián on 2–3 September 2024. Thereafter, two sets of case studies are examined: the Established AI4SI Ecosystem and the Decentralized Web3 Emerging Ecosystem, both offering different approaches to data governance and community-driven innovation. The examination of these case studies was conducted in the scope of the AI4SI International Summer School with 150 online and 100 offline participants, including scholars, policymakers, practitioners, managers, PhD students, activists, and tech entrepreneurs, among others. Ultimately, the article revolves around final remarks, limitations, and future research avenues.

2. Methods: Literature Review, Policy Analysis and Action Research Through AI4SI International Summer School

The following diagram (Figure 1) provides an overview of the paper’s methodological architecture. It clarifies how the study is structured and how its three core components—(i) the literature review, (ii) the policy analysis of the AI4SI framework, and (iii) the Action Research process conducted through the AI4SI International Summer School—connect to one another. The figure visualizes the sequential and iterative relationships between these components and demonstrates how they collectively inform the comparative assessment of the two innovation ecosystems analysed in the paper: the Established AI4SI Ecosystem and the Decentralized Web3 Emerging Ecosystem. This schematic is intended to guide the reader and strengthen the transparency of the methodological pathway followed throughout the study.

The article uses the term AI Economics to refer to the political economy of value creation, extraction, and redistribution generated specifically through machine-learning models and algorithmic infrastructures. This concept differs from digital economics, which focuses on digitalisation, platform effects, and network dynamics, and from data governance frameworks that address questions of ownership, rights, and institutional management of data. AI Economics instead concerns the economic and organisational transformations produced by predictive modelling, classification systems, and automated decision-making, as well as the broader societal and distributional implications of algorithmic systems.

This conceptualisation draws on innovation-systems theory, which treats AI as a general-purpose technology embedded in socio-technical structures rather than merely as a digital tool. At the same time, it incorporates political-economy perspectives that examine value capture, redistribution, and digital sovereignty. Together, these strands of literature help specify what is distinctively AI-driven about contemporary economic change and provide a clearer analytical foundation for comparing centralized AI4SI initiatives with decentralized Web3 social-innovation models.

The literature reveals three interconnected debates that shape contemporary discussions on AI governance and value distribution. First, analyses of large-scale biometric and data-extractive initiatives—such as WorldCoin—highlight concerns about asymmetrical power relations, consent, and the consolidation of control in transnational private actors. Second, proposals for data cooperatives and other collective data-governance mechanisms emerge from critiques of platform capitalism and seek to redistribute control over data and its derived economic value to communities. Third, the expansion of blockchain-based infrastructures and DAOs has generated debate about whether distributed governance and tokenized coordination mechanisms offer plausible alternatives to centralized AI systems. These three strands—though often discussed separately—share a common concern with addressing structural inequalities embedded in digital ecosystems and exploring institutional configurations that may enable more equitable participation in AI-driven value creation.

Situating WorldCoin, data cooperatives, and blockchain-based DAOs within a single analytical frame illustrates a broader shift: from extractive, centralized data infrastructures toward experiments in collective governance and distributed value allocation. This shift forms the conceptual space in which AI Economics can be interrogated for its potential contributions to poverty alleviation.

Despite the growing richness of these debates, there remains limited integration between discussions of AI’s political economy and the governance models that could enable redistributive outcomes in the Global South. Much of the existing literature examines AI’s risks or critiques specific technological interventions but does not link these analyses to institutional pathways capable of enhancing community agency or local control over AI-driven value flows. This paper addresses this gap by comparing centralized and decentralized innovation ecosystems through the lens of AI Economics, examining how governance architectures shape the distribution of value, resources, and decision-making power.

2.1. Literature Review: AI4SI

The rise of decentralized Web3 technologies has opened new avenues for rethinking the way data, governance, and identity management are approached, particularly in the Global South. One of the most notable projects in this space is WorldCoin [61], which aims to create a global financial network using AI and blockchain to offer cryptocurrency-based Universal Basic Income (UBI) for individuals in underserved regions. According to WorldCoin’s founders, their goal is to provide financial inclusion to billions of people who lack access to traditional banking systems. However, this approach has sparked significant debate, particularly regarding its reliance on biometric data [62]. As Gent highlights in IEEE Spectrum [63], WorldCoin’s ambition to scan the irises of the world’s population as a means of authentication raises privacy concerns, particularly in regions with fragile governance structures [64]. While WorldCoin promises a new financial reality, scholars argue that the trade-offs between privacy and access to financial resources require closer scrutiny [65].

Beyond biometric-driven initiatives such as WorldCoin, recent scholarship emphasizes that the geopolitical, institutional, and discursive contexts in which digital infrastructures are deployed play a decisive role in shaping their societal impacts [66]. As Manor and Segev illustrate [67], AI systems themselves can reinforce existing geopolitical narratives by embedding national frames and sentiments in ways that influence global digital imaginaries. Similarly, Suter et al. demonstrate that political debates on AI reveal competing issue-frames, suggesting that technologies like blockchain-based identity schemes will not be adopted in isolation but will be filtered through partisan and institutional agendas [68]. This interplay is further complicated by questions of sovereignty and governance: Repetto shows how the notion of “digital sovereignty” is contested and stretched across contexts [69], while Qu, Yuan, and Xu highlight the multi-level governance challenges in safeguarding it [70]. Taken together, these insights suggest that while projects such as WorldCoin frame themselves as neutral instruments of financial inclusion, they are embedded in wider struggles over digital governance, sovereignty, and legitimacy that extend far beyond the Global South.

Furthering the conversation on data sovereignty and community empowerment, Stein et al. present a framework for Data Cooperatives [71], which are emerging as a collective model for managing and governing data [72]. Unlike traditional centralized systems where corporations control user data, data cooperatives are designed to return control to the individual or community, enabling more equitable data governance [73,74,75]. Stein et al. argue that data cooperatives can serve as an antidote to the power imbalances perpetuated by data-opolies in the Global North, offering the Global South an opportunity to build governance systems that are inclusive, transparent, and community-driven [76,77].

In line with this, the ethical implications of decentralized technologies, as examined by Allen et al. in their work at the Harvard Edmond & Lily Safra Center for Ethics at Harvard University, explore how decentralized platforms like the Fediverse and Web3 technologies can promote digital justice [78]. They emphasize that these technologies hold promise for creating more ethical, user-centered networks, but they also stress that without proper governance frameworks, these technologies can still perpetuate existing inequalities. They advocate for a rethinking of how decentralized systems can be designed and governed to ensure that they truly serve the interests of marginalized communities, particularly in the Global South.

This concern is echoed by Rob Lalka [79], who critiques the dominance of Big Tech in Silicon Valley. He argues that the true “Venture Alchemists” of our time are those who prioritize social impact over profit, leveraging decentralized technologies to challenge the status quo and foster inclusivity, particularly in the Global South [6,12]. Lalka’s vision aligns with the focus of the AI4SI framework, which emphasizes the use of disruptive technologies to drive social innovation and digital justice globally.

Mejias and Couldry introduce the concept of data colonialism, describing the actions of Big Tech as a new form of colonial exploitation where vast amounts of personal data, often extracted from the Global South, are commodified to reinforce existing power imbalances [26]. They argue that this concentration of power within a few global corporations undermines global equity and justice, perpetuating digital inequalities that disproportionately affect marginalized communities. Mejias and Couldry advocate for collective resistance against this monopolization of data, emphasizing the need for local data sovereignty as a means of reclaiming control over digital resources [80,81,82]. They highlight how decentralized technologies, such as blockchain and data cooperatives, could offer viable alternatives to the extractive practices of Big Tech by empowering local communities to govern their data autonomously, thus fostering more equitable and just digital ecosystems. Their work underscores the urgency of shifting away from centralized models of data governance to decentralize power and ensure that digital justice is achieved globally, particularly in the Global South.

Muldoon et al. provide further insight into the ethical challenges associated with AI, particularly the hidden labor required to maintain the illusion of frictionless AI systems [60]. They reveal the grim reality of millions of data workers laboring under often appalling conditions to make AI possible, exposing the stark disparity between those who benefit from AI and those who bear its burdens. Their work advocates for SI that prioritizes worker digital rights, aligning with the broader discourse on ensuring that AI technologies are not only efficient but also ethical and just.

Blockchain technology has also been identified as a key tool in the push toward decentralization. De Filippi et al. recently explore how blockchain challenges traditional legal and political frameworks, offering new paradigms of governance by decentralizing authority and redefining concepts like sovereignty and legality [77]. Blockchain’s ability to create “trustless” systems—where code, rather than centralized institutions, governs behavior—represents a significant innovation in governance, particularly in regions where traditional governance structures may be weak or corrupt. This shift toward decentralized governance has implications for how AI technologies can be deployed in a fair and just manner in the Global South to potentially alleviate poverty.

The potential of DAOs to empower marginalized communities is another area of exploration. Van Kerckhoven and Chohan examine how DAOs provide a model for decentralized decision-making that allows communities to self-govern without reliance on centralized authorities [83]. Despite their potential, DAOs come with inherent vulnerabilities, such as security risks and governance challenges. Van Kerckhoven and Chohan discuss how these vulnerabilities can be mitigated, ensuring that decentralized technologies are robust enough to serve the needs of the population, particularly in regions where traditional governance structures have failed.

The broader challenge of ethical governance in digital spaces is also explored by Schneider [84], who advocates for creating online environments that are democratically governed. He argues that digital spaces must balance technological innovation with ethical governance, emphasizing fairness, inclusivity, and user autonomy. His exploration of embedding democratic principles in the design of online spaces offers valuable insights for reimagining how AI systems are governed, ensuring that they promote equitable outcomes for all users, especially in the Global South.

Finally, scholars like Eubanks [85] and D’Ignazio and Klein [86] delve into how AI and high-tech tools, often designed without consideration for their broader social impact, can exacerbate inequalities. Eubanks criticizes how data-driven systems profile, police, and punish marginalized communities, particularly the poor, while D’Ignazio and Klein challenge the predominantly white, male-dominated narratives in AI and data science. They argue for a more intersectional approach to data ethics, one that considers the needs and rights of vulnerable populations and promotes more equitable technological solutions.

In conclusion, the literature on AI4SI highlights the complex interplay between AI, decentralized technologies, and SI. Scholars agree that while these technologies hold immense potential for fostering inclusivity and digital justice, they must be carefully governed to avoid exacerbating existing inequalities, particularly in the Global South [12]. As the debate around AI’s role in addressing systemic disparities continues, these decentralized frameworks offer a critical path forward for creating more equitable and inclusive digital futures [87]. Collectively, these perspectives provide a multifaceted view of how AI and decentralized technologies can contribute to more equitable governance structures, financial inclusion, and ethical data management in the Global South. As the debate surrounding WorldCoin, data cooperatives, and decentralized social technologies unfolds, the potential for these innovations to alleviate poverty while safeguarding data sovereignty and human rights becomes a critical focal point for future research and policy development.

2.2. Policy Analysis: AI4SI Framework

The AI4SI framework, as introduced by the World Economic Forum [9,33,34,35,44], presents a policy approach to using AI as a tool for addressing critical societal challenges, including poverty in the Global South. The WEF, through its AI for Social Impact initiative launched at the Davos Annual Meeting, has laid the groundwork for understanding how AI can be applied in SI across various sectors [36]. However, the framework’s potential in effectively addressing these challenges—particularly in lower-income regions—requires critical analysis, especially when considering the digital divide, governance structures, and the socio-political landscape of the Global South, and in light of this article, the missing perspective of decentralization and the potential role of Web3 emerging technologies [88,89,90,91,92,93,94,95,96,97,98,99,100].

The WEF’s AI4SI initiative has conducted extensive mapping of SI sectors where AI is being deployed, identifying over 300 social innovators across 50 countries and over 90 key initiatives [9]. The focus areas span from healthcare and environmental sustainability to economic empowerment. This mapping highlights the transformative potential of AI in sectors like healthcare, where 25% of innovators are using AI to improve access to medical services, and economic empowerment, where AI is helping to enhance livelihoods, particularly in low- and middle-income countries. According to the WEF’s reports, nearly 80% of AI-based economic empowerment initiatives are concentrated in these regions.

However, despite the promise of AI in these areas, the WEF findings also reveal significant disparities between the Global North and South in terms of AI adoption, infrastructure, and support for social innovators. While high-income countries like the United States and those in Europe possess robust AI ecosystems, the Global South lacks the necessary digital infrastructure and institutional support to fully harness AI for SI. For instance, the adoption of machine learning (ML), which is a key AI technology, is notably lower in Africa (50%) and Oceania (40%) compared to the global average of 70%. This disparity is further exacerbated by challenges such as limited access to high-quality data, algorithmic biases, and inadequate regulatory frameworks.

The challenges identified by the WEF [9,33,34,35], particularly in the Global South, underscore the need for a tailored policy approach to AI implementation. One of the key issues highlighted in the AI4SI framework is the digital divide, which manifests not only in access to technology but also in the availability of resources and expertise to effectively deploy AI solutions [101]. The WEF reports point to a significant gap in AI skills and digital literacy, particularly among rural and marginalized communities. This disparity raises concerns about the scalability of AI-driven solutions and their ability to reach the communities most in need [102].

In addition to infrastructure challenges, the WEF identifies several ethical dilemmas related to AI deployment, such as data privacy, biases in AI systems, and the need for ethical oversight. While AI has the potential to revolutionize sectors like healthcare and education, it can also exacerbate existing inequalities if not implemented with a focus on fairness and inclusivity. The WEF’s PRISM framework for responsible AI emphasizes the importance of building trust, ensuring transparency, and fostering collaboration between AI developers, policymakers, and communities to address these ethical concerns [33]. However, the framework does not sufficiently address the structural power imbalances created by centralized data governance models, often perpetuated by large corporations based in the Global North [44].

One of the critiques of the WEF’s approach to AI4SI is its underemphasis on decentralized technologies, such as blockchain, which could provide a solution to the data governance challenges identified in the Global South [103]. Decentralization, as explored through AI4SI’s International Summer School and recent scholarship, presents a significant opportunity for reshaping how data and AI are governed, particularly in low- and middle-income countries [12]. By leveraging decentralized technologies such as blockchain, DAOs [104], and data cooperatives, the Global South could potentially overcome some of the infrastructural and governance challenges that inhibit the effective implementation of AI.

Decentralized models of governance offer a way to distribute power and decision-making authority more equitably, bypassing traditional top-down governance structures that often fail to serve marginalized communities [93]. This is particularly relevant in regions where governance structures are fragile or corrupt, as decentralized systems reduce the need for reliance on centralized authorities and instead promote peer-to-peer networks and community-driven governance.

The WEF’s PRISM framework [33], which advocates for responsible AI, could benefit from incorporating these decentralized governance models into its strategy [105]. By doing so, the framework would not only address the immediate technical and ethical concerns surrounding AI but also promote long-term solutions that empower local communities and foster greater data sovereignty [106,107,108]. Moreover, blockchain’s capacity to create “trustless” systems could help mitigate issues related to data privacy and algorithmic bias, which are critical concerns for social innovators operating in the Global South [109,110,111].

In conclusion, while the WEF’s AI4SI framework presents a promising roadmap for leveraging AI to address poverty and other social challenges, its effectiveness in the Global South depends on addressing several key issues [112]. First, the digital divide must be bridged through targeted investments in digital infrastructure and AI education. Second, the framework needs to integrate decentralized technologies like blockchain [109], which offer the potential to democratize data governance and empower marginalized communities. Finally, ethical oversight and a focus on digital justice must be at the core of any AI4SI initiative to ensure that AI-driven solutions are both equitable and sustainable in the long run. By embedding these principles into the AI4SI policy framework, an Action Research approach can help create a more inclusive and just digital future for the Global South, ensuring that AI serves as a tool for poverty alleviation and SI rather than exacerbating existing inequalities [13,38,39,113,114].

Balancing the Critique of the WEF’s AI4SI and PRISM Frameworks: The WEF’s AI4SI and its PRISM Framework for Responsible AI make significant contributions to establishing ethical and inclusive standards for AI governance. PRISM’s emphasis on Purpose, Responsibility, Integrity, Sustainability, and Manageability sets a strong normative baseline that foregrounds trust, transparency, and multi-stakeholder engagement. These features constitute valuable advances toward equitable AI adoption, especially in contexts lacking regulatory maturity.

The critique advanced in this article does not dispute these achievements but highlights a missing layer of structural decentralization. While PRISM successfully delineates the “ethical what,” it remains largely silent on the “institutional how” of power redistribution in data governance. Integrating decentralization would operationalize PRISM’s commitments by embedding participatory governance mechanisms within its pillars:

(i)

Under Responsibility, data cooperatives could formalize collective consent and accountability.

(ii)

Under Integrity, blockchain-based audit trails could enhance transparency in algorithmic processes.

(iii)

Under Manageability, DAOs could enable adaptive oversight through community-driven rule-making.

In this sense, decentralization is not a competing paradigm but a complementary governance layer that can translate PRISM’s ethical intentions into enforceable, participatory practice. A hybrid integration of PRISM and decentralized governance would allow the AI4SI framework to couple its existing ethical rigour with mechanisms for distributed decision-making and local data sovereignty [9,33,34,35,44,105,106,107,108].

2.3. Action Research Through AI4SI International Summer School

The 2024 AI4SI International Summer School, held in Donostia-San Sebastián (Spain) on 2 and 3 September 2024, employed Action Research (AR) as its core methodology, allowing for the real-time application of theoretical insights and the co-creation of knowledge. AR, a participatory approach rooted in iterative cycles of planning, action, observation, and reflection, has been instrumental in bridging the gap between research and practice, particularly in contexts like the Global South [13]. By engaging participants from diverse backgrounds—scholars, policymakers, practitioners, activists, and PhD students—this event fostered an environment of collective inquiry where theoretical frameworks such as AI4SI were tested and refined based on local needs and global challenges. The AI4SI 2024 International Summer School engaged participants through several dynamics, blending speakers and participants around different questions and tasks, either offline or online.

AR is particularly suited to the AI4SI framework, given the complex nature of integrating emerging technologies such as AI and decentralized Web3 frameworks into rural and urban development strategies in the Global South [114]. This approach allowed 100 offline and 150 online participants representing 15 countries (Costa Rica, Mexico, Brazil, Guatemala, Peru, Colombia, Kenya, Mozambique, South Africa, Malawi, Spain, the USA, Belgium, the UK, and Switzerland) to work directly with the technology, confronting real-world challenges and refining solutions in an iterative process. The AI4SI International Summer School exemplified this by engaging participants in hands-on workshops, collaborative discussions, and scenario planning exercises that focused on applying AI and decentralized technologies like blockchain and DAOs to address issues of poverty and inequality in marginalized communities. Following the good results of this AR fieldwork in 2024, in July 2025, the Province Council of Gipuzkoa has held under the same scientific direction of the author of this article, another International Summer School entitled Digital Inclusion & GenAI: https://www.uik.eus/en/activity/digital-inclusion-generative-artificial-intelligence-gipuzkoa-socially-cohesive-digitally (accessed on 1 November 2025). The main results of this international summer school has been published in Open Access in the journal Discover Cities [87].

In September 2024 International Summer School, the AR dynamic focused on how participants approached digital justice (definition, opinion, proximity, prospective, and projective perspectives) through different contexts (daily, awareness, critical factors, employment, sustainability, gender, youth, literacy, decentralization, and present/future). Alongside this empirical approach, participants gradually identified and brainstormed case studies that the dynamic gathered as the main result collected for this article. Consequently, after this generic introduction to digital justice and its multiple contextualizations, one of the key outcomes of the AR process at the AI4SI International Summer School was the identification and critical examination of case studies that focused on inclusivity and community-driven development. These case studies played an essential role in addressing the central research question of this article: Can the emerging field of AI Economics—the study of how AI redistributes value, labor, and resources—contribute to poverty alleviation in the Global South? [115,116]. As a result of this process, two distinct groups of case studies emerged:

(i)

AI4SI Established Ecosystem: This group includes case studies that, in their original policy formulations, emphasize how AI can be applied for SI. However, these cases typically do not question the nature of data governance, nor do they consider whether the architecture of these solutions is centralized or decentralized.

(ii)

AI4SI Decentralized Web3 Emerging Ecosystem: In contrast, this second group includes case studies that not only address AI for SI purposes but also align with principles of decentralization. These cases specifically focus on decentralized technologies such as blockchain and Web3 [12].

This examination was built upon the principle that technological solutions must not only address immediate needs but also empower communities to take ownership of their data and governance processes. For example, participants explored how data cooperatives could serve as a decentralized model for data governance [66,71,72,73,74,75,117,118,119,120]; i.e., www.salus.coop (accessed on 1 November 2025), enabling communities in the Global South to manage and control their data autonomously, rather than relying on external corporations or centralized authorities. This aligns with the broader goals of digital justice, which seek to redress power imbalances in the digital economy.

Moreover, the iterative nature of AR allowed participants to identify potential ethical and practical challenges associated with deploying AI in the Global South. These challenges included algorithmic biases, data privacy concerns, and the risk of exacerbating existing inequalities [121]. By reflecting on these challenges in real time, participants could propose solutions that were context-sensitive and ethically grounded [37]. One notable insight was the recognition that AI technologies, when introduced without adequate local engagement, risk reinforcing existing power structures rather than dismantling them [5]. As such, the AR process emphasized the importance of involving local communities not just as beneficiaries of technological innovation but as active participants in shaping how these technologies are used [52].

The AI4SI International Summer School also highlighted the potential of decentralized technologies to foster more equitable governance models [78]. Through AR, participants were able to experiment with the use of blockchain and DAOs to facilitate decentralized decision-making processes, particularly in regions with fragile governance systems. These technologies, which allow for peer-to-peer interactions and reduce reliance on centralized authorities, were seen as critical tools for promoting transparency, accountability, and community ownership in development efforts. The integration of these technologies into the AI4SI case studies provided a pathway for addressing the governance challenges that often hinder the effective deployment of AI in the Global South.

Yet, as Calzada cautions, the apparent promise of decentralized governance through Web3 technologies can mask deeper contradictions that risk reproducing existing asymmetries rather than dismantling them [122]. While blockchain and DAOs may rhetorically advance ideals of transparency and collective empowerment, in practice they often consolidate control in the hands of technologically literate elites, thereby creating new forms of exclusion. Moreover, the narratives of decentralization frequently overlook structural issues of power, geopolitics, and institutional fragility, particularly acute in the Global South. This suggests that without careful regulation, participatory safeguards, and contextual sensitivity, the deployment of decentralized infrastructures may evolve less as emancipatory tools for democratic governance and more as instruments that reinforce digital dependency and techno-solutionist imaginaries.

The AR process followed a structured, iterative design aligned with established AR frameworks. It unfolded across two main cycles. Cycle 1 involved online engagement with approximately 150 participants through a series of thematic workshops focusing on digital inclusion, AI-enabled social innovation frameworks, and decentralization. Each workshop included short inputs, facilitated dialogue, and small-group reflections, which generated preliminary insights into the governance and value-creation challenges in AI4SI initiatives. Cycle 2 was conducted offline during the AI4SI International Summer School, where around 100 participants engaged in participatory scenario exercises, case-study mapping, and group-based problem-solving activities. This second cycle enabled the refinement of concepts initially generated in Cycle 1 and provided opportunities to observe how participants interpreted both centralized and decentralized models within real-world contexts.

Across both cycles, data collection followed a qualitative, non-intrusive protocol. The author gathered anonymized observational notes, workshop artefacts (e.g., flipcharts, collaborative boards, group summaries), participant reflections, and thematic outputs generated during the exercises. No personally identifiable information was recorded. These materials informed the identification of governance patterns, redistribution mechanisms, and perceived opportunities and constraints in both ecosystems.

Case studies included in the Established and Emerging Ecosystems were selected using relevance sampling, based on three criteria: (i) documented use of AI or Web3 technologies for social-innovation purposes, (ii) clear articulation of governance or value-distribution mechanisms, and (iii) presence of publicly accessible information enabling desk-based validation. These case studies were not co-constructed with participants but were used as analytical inputs during group work to stimulate comparative reflection. Their inclusion in

Table 1. AI4SI Established Ecosystem: 25 Case Studies.

Case Study	Sector	HQ	Link	Indicative Metrics (Qualitative Severity Scale)
				Governance Openness	Redistribution Potential	Technical Complexity/Barrier to Entry	Institutional Maturity	Risk of Elite Capture
1. DIMAGI	Deploys chatbots for evidence-based healthcare in low-resource settings.	USA	https://www.dimagi.com/ (accessed on 1 November 2025)	Low	Medium	Medium	High	Medium-High
2. LifeBank	Uses AI to improve healthcare delivery in underserved regions.	Nigeria	https://www.lifebank.ng/ (accessed on 1 November 2025)	Low	Medium	Medium	High	Medium
3. High Resolves	Focuses on overcoming connectivity challenges to deploy AI in education and social initiatives.	Australia	https://highresolves.org/ (accessed on 1 November 2025)	Low	Low-Medium	Medium	Medium	Medium
4. RECODE	A partnership promoting impactful AI through social innovation.	Brazil	https://recode.org.br/ (accessed on 1 November 2025)	Low	Medium	Medium	Medium	Medium
5. Education for Employment	Deploys AI to enhance operational efficiency in education and employment programs.	MENA	https://www.efe.org/ (accessed on 1 November 2025)	Low	Low–Medium	Medium	High	Medium
6. Ashoka’s Tinkering Approach	Implements AI internally within Ashoka to enhance social innovation processes.	USA	https://www.ashoka.org/ (accessed on 1 November 2025)	Low	Low	Medium	High	Medium-High
7. SAS Brasil’s Cervical Cancer Screening	Uses AI for early cervical cancer screening in underserved areas.	Brazil	https://www.sasbrasil.org.br/ (accessed on 1 November 2025)	Low	Medium	High	Medium-High	Medium
8. Geekie	Provides AI-powered personalized education tools for students.	Brazil	https://www.geekie.com.br/ (accessed on 1 November 2025)	Low	Medium	Medium-High	High	Medium-High
9. TensorFlow Lite	A lightweight AI framework for deploying machine learning models on mobile devices and low-resource environments.	USA	https://www.tensorflow.org/lite (accessed on 1 November 2025)	Low	Medium	High	High	Medium-High
10. Access Earth	A platform for finding and rating places based on accessibility, promoting inclusivity.	Ireland	https://accessearth.com/ (accessed on 1 November 2025)	Medium	Medium	Medium	Medium	Medium
11. Agrimetrics	Provides agricultural data analytics to improve crop value chains and food security.	UK	https://app.agrimetrics.co.uk/ (accessed on 1 November 2025)	Low	Medium	High	High	Medium-High
12. BetterSpace	An employee wellbeing platform offering tools for mental health and productivity.	UK	https://www.betterspace.uk/ (accessed on 1 November 2025)	Low	Low-Medium	Medium	High	Medium
13. Citymaas	Provides accessibility information for travelers, enhancing mobility for individuals with disabilities.	UK	https://www.citymaas.io/ (accessed on 1 November 2025)	Low	Medium	Medium	Medium	Medium
14. ev.energy	Manages electric vehicle charging, optimizing for sustainability and cost-efficiency.	UK	https://www.ev.energy/ (accessed on 1 November 2025)	Low	Low	High	High	Medium-High
15. iDyslexic	A social app connecting individuals with dyslexia and ADHD for support and secure communication.	Ireland	https://www.crunchbase.com/organization/idyslexic (accessed on 1 November 2025)	Medium	Medium	Low-Medium	Medium	Medium
16. Immersive Rehab	Utilizes virtual reality for improving physical and neuro-rehabilitation outcomes.	UK	https://www.immersiverehab.com/ (accessed on 1 November 2025)	Low	Medium	High	Medium	Medium
17. ThermaFY	Offers thermal imaging software for healthcare and medical research applications.	UK	https://www.thermafy.com/ (accessed on 1 November 2025)	Low	Low-Medium	Medium	Medium	Medium
18. Upstream Health	Provides innovative technologies for health and social care teams.	Turkey	https://www.upstream.health/ (accessed on 1 November 2025)	Low	Low-Medium	Medium	Medium	Medium
19. WeWalk	Develops smart canes and apps to improve mobility for visually impaired individuals.	South Africa	https://wewalk.io/en/ (accessed on 1 November 2025)	Medium	Medium	Medium	Medium	Medium
20. Africa TeenGeeks	Empowers African youth through coding and digital skills education.	Nigeria	https://www.africateengeeks.co.za/ (accessed on 1 November 2025)	Medium	High	Low-Medium	Medium	Medium
21. Youth for Technology Foundation	Provides technology-based education and empowerment programs for youth in underserved communities.	Kenya	https://www.youthfortechnology.org/ (accessed on 1 November 2025)	Medium	High	Medium	Medium	Medium
22. BRCK	Offers internet connectivity solutions in remote areas to promote digital inclusion.	USA	https://www.brck.com/ (accessed on 1 November 2025)	Low	Medium	Medium	Medium	Medium
23. Suki.ai	AI-powered medical assistant that streamlines documentation for healthcare providers.	Brazil	https://www.suki.ai/ (accessed on 1 November 2025)	Low	Medium	High	High	High
24. MapBiomas	Uses AI for environmental monitoring and management in Brazil.	Kenya	https://plataforma.brasil.mapbiomas.org/ (accessed on 1 November 2025)	Low	High	High	Medium	Medium
25. Apollo Agriculture	Provides data-driven insights to improve farming productivity and sustainability.	Kenya	https://www.apolloagriculture.com/ (accessed on 1 November 2025)	Low	Medium	Medium-High	Medium	Medium

and

Table 2. AI4SI Decentralized Web3 Emerging Ecosystem: 14 Case Studies.

Case Study	Sector	HQ	Link	Indicative Metrics (Qualitative Severity Scale)
				Governance Openness	Redistribution Potential	Technical Complexity/Barrier to Entry	Institutional Maturity	Risk of Elite Capture
1. Fishcoin	A blockchain-based platform incentivizing sustainable seafood supply chain management.	Singapore	https://fishcoin.co/ (accessed on 1 November 2025 )	High	High	High	Medium	Medium-High
2. Commons Stack	Uses DAOs to create sustainable digital ecosystems and commons.	Switzerland	https://commonsstack.org/ (accessed on 1 November 2025)	High	High	High	Medium	Medium
3. Gitcoin	A DAO that funds open-source software development, focusing on community-driven innovation.	USA	https://gitcoin.co/ (accessed on 1 November 2025)	High	High	High	High	High
4. Grassrootseconomics.org	Utilizes DAOs to support community currency projects in marginalized areas, empowering local economies.	Kenya	https://www.grassrootseconomics.org/ (accessed on 1 November 2025)	High	High	Medium-High	Medium-High	Medium
5. AID	Provides decentralized aid distribution using blockchain to ensure transparency and efficiency.	Global	N/A	Medium-High	High	High	Low-Medium	Medium
6. Grassroots Economics	A social enterprise using blockchain and DAOs to implement community inclusion currencies in Africa.	Kenya	https://www.grassrootseconomics.org/ (accessed on 1 November 2025)	High	High	Medium	Medium-High	Medium
7. BanQu	A blockchain-based platform aimed at providing transparency and traceability in supply chains, focusing on marginalized producers.	USA	https://banqu.co/ (accessed on 1 November 2025)	Medium	High	High	Medium-High	Medium-High
8. Amply	A blockchain-based platform that uses self-sovereign identity to verify and incentivize early childhood development programs.	South Africa	http://amply.tech/ (accessed on 1 November 2025)	Medium-High	Medium-High	High	Medium	Medium-High
9. Kenyan AI and Blockchain Taskforce	A government initiative exploring AI and blockchain applications to drive innovation and economic development in Kenya.	Kenya	N/A	Medium	Medium	Medium	High	Medium
10. Uniswap Grants Program	A decentralized exchange supporting projects that promote diversity, inclusion, and accessibility within the DeFi ecosystem.	USA	https://unigrants.org/ (accessed on 1 November 2025)	Medium-High	Medium-High	High	High	High
11. Giveth	A decentralized platform enabling transparent charitable donations using blockchain technology.	Spain	https://giveth.io/ (accessed on 1 November 2025)	High	High	Medium-High	Medium	Medium
12. Consensys Social Impact	Collaborates with NGOs to leverage blockchain for social impact, focusing on areas like disaster relief and supply chain transparency.	USA	https://consensys.net/ (accessed on 1 November 2025)	Medium	Medium-High	High	High	Medium-High
13. Powerledger	A blockchain-based energy trading platform that enables decentralized energy markets.	Australia	https://www.powerledger.io/ (accessed on 1 November 2025)	Medium-High	Medium	High	High	Medium-High
14. WePower	A blockchain-based platform enabling green energy trading, focusing on connecting producers and consumers directly.	Lithuania	https://wepower.network/ (accessed on 1 November 2025)	Medium-High	Medium-High	High	Medium	Medium-High

is therefore grounded in desk research triangulated with insights from the AR cycles.

Ethical approval for the AR activities was obtained through the institutional procedures of the author’s host institution. All participants provided informed consent, and participation was voluntary. The activities were designed to avoid risks, and no sensitive personal data were collected.

The themes that informed the policy recommendations emerged from systematic analysis of workshop artefacts—including group reflections, co-created governance maps, participatory scenario exercises, and recorded thematic outputs. These materials provided empirical grounding for identifying governance shortcomings, capacity gaps, and areas where centralized and decentralized models require institutional strengthening.

In conclusion, the AR methodology employed during the AI4SI International Summer School allowed for a dynamic and responsive exploration of how AI and decentralized technologies could be harnessed for SI. By fostering a participatory and iterative process, the event enabled participants to co-create solutions that were both contextually relevant and globally informed, ensuring that AI-driven initiatives promote inclusivity, digital justice, and equitable development in the Global South. This AR-based approach offered a valuable model for future initiatives seeking to integrate technology with social impact in a way that is both ethical and sustainable. As such, the next section discusses the two groups of cases identified: (i) AI4SI Established Ecosystem and (ii) AI4SI Decentralized Web3 Emerging Ecosystem.

3. Results: AI4SI Established and Emerging Ecosystems of Case Studies

To ensure transparency and methodological rigour, the five qualitative metrics introduced in Table 1 and Table 2 were derived directly from the article’s conceptual framework on AI Economics, the comparative logic of the two ecosystems, and the insights generated through the Action Research conducted during the AI4SI International Summer School. The metrics operationalize central dimensions discussed throughout the paper—governance, redistribution, technical capacity, institutional readiness, and risks of concentration—and translate them into a structured comparative format consistent with the study’s qualitative design.

• Governance Openness

Derived from the paper’s emphasis on centralized vs. decentralized data governance models, this metric evaluates the degree of participatory decision-making, transparency, and community involvement reflected in each case. It aligns with debates on digital sovereignty and data cooperatives discussed in the literature review.

2.

Redistribution Potential

This metric reflects the core focus of AI Economics on value flows, power asymmetries, and the possibility of redistributing economic or informational benefits. It closely follows the article’s analytical distinction between service-delivery AI and redistributive Web3 models.

3.

Technical Complexity/Barrier to Entry

Technical sophistication repeatedly emerges in the paper as a major determinant of exclusion, digital divides, and risks of elite capture. This metric operationalizes the extent to which each case study requires specialised skills, infrastructure, or costly technological integration, echoing the discussions in Section 2 and Section 4 as well as the concerns raised in the AR workshops.

4.

Institutional Maturity

This parameter reflects the stability, longevity, and organisational consolidation of each initiative. It is grounded in the comparative analysis between well-established AI4SI projects and the more experimental Web3 emerging ecosystem described in Section 3.

5.

Risk of Elite Capture

A central theme in the paper—supported by both the literature and the AR findings—is the risk that both centralized AI systems and decentralized Web3 ecosystems may reinforce inequalities. This metric captures the likelihood that benefits, data, or decision-making could be monopolized by digital, technical, or institutional elites.

• Why a Qualitative Severity Scale?

Given the heterogeneity of sectors, geographies, and models represented in the AI4SI case studies, quantitative indicators (e.g., numerical performance metrics) would be neither comparable nor meaningful across all initiatives. Instead, a standardized qualitative scale (Low, Medium, High) allows for consistent cross-case interpretation while respecting the qualitative nature of the policy analysis and Action Research methodology.

• Purpose of Adding Metrics

The metrics do not claim causal inference; instead, they provide:

• A systematic analytical structure requested by the reviewers
• A semi-quantitative complement to narrative findings
• A way to make explicit the normative dimensions—governance, equity, capability, sovereignty—that underpin the article’s argument

Together, these metrics increase the explanatory clarity of Table 1 and Table 2 and bring conceptual coherence to the comparison between the Established AI4SI Ecosystem and the Decentralized Web3 Emerging Ecosystem.

The initiatives listed in Table 1 and Table 2 were identified through purposeful relevance sampling, designed to capture illustrative examples of AI-enabled social-innovation initiatives across diverse geographical and sectoral contexts. The objective was not to generate a statistically representative dataset but to assemble comparative cases that exhibit clear differences in governance structures, data architectures, and value-distribution mechanisms. Three criteria guided inclusion: (i) documented use of AI or Web3 technologies for social innovation; (ii) availability of publicly accessible information enabling desk-based validation; and (iii) relevance to the study’s focus on governance, redistribution, and digital inclusion.

The classification of initiatives into the “Established” and “Emerging” Ecosystems is conceptual rather than chronological. The Established Ecosystem refers to AI4SI initiatives operating within centralized data and decision-making infrastructures, even when the initiatives themselves are relatively new. Conversely, the Emerging Ecosystem consists of initiatives grounded in decentralized architectures—blockchain, DAOs, and self-sovereign identity—regardless of their maturity or age. This approach explains why some recent projects such as BetterSpace or WeWalk are categorized as part of the Established Ecosystem: despite their innovative contributions, they rely on centralized governance structures and do not incorporate decentralization, tokenized governance, or community-managed data stewardship.

3.1. AI4SI Established Ecosystem

The AI4SI Established Ecosystem focuses on sectors like healthcare, education, agriculture, and accessibility. These initiatives span different regions globally and aim to tackle critical issues such as healthcare access, education inequality, and environmental sustainability. The case studies presented in Table 1 offer a glimpse into the wide variety of cases in which AI is being applied to address social challenges in both the Global North and South. However, one commonality is that these initiatives often rely on centralized data governance models, raising questions about the long-term sustainability and inclusivity of these solutions.

For example, DIMAGI, based in the USA, deploys AI-powered chatbots for healthcare in low-resource settings. While the initiative is transformative in terms of improving healthcare delivery, it operates within a centralized framework where control over data and technological infrastructure remains with a single entity. Similarly, LifeBank in Nigeria utilizes AI to optimize healthcare delivery by ensuring the timely distribution of critical medical supplies. Although highly impactful, LifeBank, too, lacks a focus on decentralization or local data sovereignty.

Another case, RECODE in Brazil, partners with local entities to use AI for SI, especially in education. However, much like the others in the Established Ecosystem, it does not delve into the governance structure of data and decision-making processes, which remain centralized. Geekie, another Brazilian initiative, uses AI to provide personalized education tools, but the model operates within a traditional, centralized framework, meaning local communities have limited control over how the data is used.

The AI4SI Established Ecosystem excels in scaling AI solutions to address urgent social issues. Yet, despite their potential feasibility in sectors like healthcare, agriculture, and education, these initiatives do not fundamentally challenge existing power structures or data governance models. Most projects rely on centralized platforms, which, while efficient, may limit long-term community ownership and perpetuate dependence on external actors for technological solutions. As AI-driven SI continues to evolve, there remains an opportunity to explore how decentralization could enhance inclusivity, transparency, and long-term sustainability.

The Indicative Metrics column offers a qualitative severity scale that systematizes variation across governance openness, redistribution potential, technical complexity, institutional maturity, and risk of elite capture. This semi-quantitative layer complements the narrative discussion by providing a structured comparative assessment of the centralised AI4SI Established Ecosystem.

3.2. AI4SI Decentralized Web3 Emerging Ecosystem

In contrast, the AI4SI Decentralized Web3 Emerging Ecosystem, as showcased in Table 2, represents a shift toward the use of decentralized technologies such as blockchain and DAOs to drive SI. These case studies not only leverage AI but also incorporate decentralized governance models, offering a more community-driven and transparent opportunities to addressing social challenges [122,123,124]. Unlike the Established Ecosystem, these emerging initiatives aim to redistribute power and data control back to local communities, ensuring that the benefits of AI and Web3 technologies are shared more equitably.

The AI4SI Decentralized Web3 Emerging Ecosystem refers to a new class of social innovation initiatives that merge AI with Web3 technologies—blockchain, DAOs, and data cooperatives—to enable equitable, transparent, and data-sovereign development models. In this ecosystem, AI acts as a computational intelligence layer within decentralized infrastructures. Rather than serving merely as a data analytics add-on, AI algorithms actively process on-chain and off-chain data to support predictive modeling, anomaly detection, dynamic governance, and context-sensitive decision-making. For example, Fishcoin uses AI-driven anomaly detection to flag irregularities in seafood supply-chain data and applies predictive analytics to forecast sustainable fishing quotas; Grassroots Economics applies AI to model transaction patterns in community currencies, improving network stability; Commons Stack and Gitcoin use machine-learning models to evaluate project impact and guide DAO resource allocation; BanQu integrates AI for real-time identity verification and traceability of producers; and Giveth employs natural-language-processing algorithms to classify and prioritize community proposals. Through these applications, AI transforms decentralized systems from static ledgers into adaptive socio-technical ecosystems capable of collective learning, prediction, and governance optimization. This integration defines the distinct contribution of AI within the broader Web3 social innovation paradigm [12,77,83].

One standout example is Fishcoin, a blockchain-based platform incentivizing sustainable seafood supply chains. By using decentralized technology, Fishcoin ensures transparency throughout the supply chain, empowering local producers and consumers alike. This approach contrasts sharply with traditional centralized models, where control often rests with a few powerful actors. Similarly, Commons Stack uses DAOs to create sustainable digital ecosystems, providing a participatory framework for communities to co-manage resources. Such initiatives demonstrate the potential of decentralized governance to not only optimize resource distribution but also foster digital justice and inclusivity.

Another key player in the Emerging Ecosystem is Grassroots Economics, which uses DAOs and blockchain to support community currency projects in marginalized areas of Kenya. This project highlights how decentralization can empower local economies by allowing communities to create and manage their own financial systems without relying on traditional banking infrastructure. Furthermore, platforms like Giveth and Gitcoin leverage blockchain to ensure transparency in charitable donations and open-source software funding, respectively, promoting community-driven innovation and reducing dependency on centralized financial systems.

These case studies from the AI4SI Decentralized Web3 Emerging Ecosystem not only seem to address immediate social challenges but also might offer a path toward long-term sustainability by decentralizing power and ensuring greater community involvement in governance. By moving away from traditional, hierarchical models, these initiatives might embody the core principles of digital justice and could ensure that the benefits of AI and Web3 technologies are distributed more equitably and transparently across all participants.

The Indicative Metrics columns provide a structured qualitative comparison of decentralized Web3 initiatives across five analytical dimensions aligned with the article’s conceptual framework: governance openness, redistribution potential, technical complexity, institutional maturity, and risk of elite capture. These values reflect qualitative judgments based on each initiative’s governance architecture, technological configuration, and operational model as described in Section 3.2.

In the Decentralized Web3 Emerging Ecosystem, AI operates as an embedded intelligence layer that enables learning, prediction, and adaptive governance. For instance, in Fishcoin, AI algorithms analyze supply-chain data to verify sustainability claims and forecast demand variations, improving market efficiency. Grassroots Economics employs AI-driven analytics to model transaction flows in community currencies and detect anomalies that could undermine trust. In Commons Stack and Gitcoin, machine learning models are used to evaluate project performance, optimize token allocation, and support deliberative decision-making within DAOs. Similarly, BanQu integrates AI-powered verification systems to authenticate digital identities and trace the provenance of goods, while Giveth applies natural language processing (NLP) to classify and rank community proposals transparently. In each of these cases, AI provides the cognitive infrastructure that transforms decentralized systems from static ledgers into adaptive, data-informed environments capable of supporting dynamic and equitable governance. By combining AI’s predictive and analytical capacities with Web3’s decentralized architectures, these models begin to illustrate how socially oriented AI can be operationalized in distributed, community-driven innovation ecosystems [12,77,83].

In conclusion, the key distinction between the Established and Emerging Ecosystems lies in their approach to governance and community involvement. The AI4SI Established Ecosystem represents a collection of platforms that effectively deploy AI solutions to address critical social challenges, such as healthcare, education, and environmental sustainability, across diverse geographies like the USA, Brazil, Nigeria, and the UK. While these initiatives might succeed in scaling impactful solutions, they typically rely on centralized data governance models, limiting local communities’ control over their data and decision-making processes. For instance, platforms like DIMAGI (Cambridge, MA, USA) and LifeBank (Lagos, Nigeria) demonstrate clear social benefits but remain within a centralized framework, which raises concerns about long-term sustainability and community ownership.

In contrast, the AI4SI Decentralized Web3 Emerging Ecosystem emphasizes decentralization through blockchain, DAOs, and self-sovereign identity models. Platforms like Commons Stack (Zug, Switzerland) and Grassroots Economics (Kilifi, Kenya) prioritize community-driven innovation and redistribute power, offering more inclusive governance models. This ecosystem shows a promising shift toward empowering local communities, ensuring transparency, and providing long-term sustainability by enabling participants to co-manage digital ecosystems.

However, it remains to be seen how both ecosystems will evolve. The Established Ecosystem holds the advantage of scalability and immediate applicability, yet it risks reinforcing existing inequalities if the centralized structures continue to dominate. On the other hand, the Emerging Ecosystem, though nascent and more experimental, promises a transformative approach to social innovation, particularly in the Global South. As AI4SI continues to develop, the integration of decentralized principles could provide a pathway to achieving digital justice and more equitable resource distribution to gradually alleviate poverty in the Global South. Whether these ecosystems can be harmonized or will continue to diverge is a question that future iterations of AI4SI must address.

4. Discussion: Responding the Research Question Through a Critical Action Research Perspective

AI has been positioned as a transformative tool in global development, but its contribution to poverty alleviation in the Global South hinges on the broader political economy of value creation, extraction, and redistribution—what this article has termed AI Economics. The AR process of the AI4SI International Summer School, along with the comparative analysis of case studies across the Established and Emerging Ecosystems (Table 1 and Table 2), provides rich insights into this question. In this discussion, we critically examine how AI Economics unfolds in practice, identifying conditions under which AI can advance social inclusion and contexts where it risks entrenching dependency and inequality.

The comparative framework therefore distinguishes ecosystems based on governance architectures rather than project age, scale, or geographical distribution, resulting in a non-representative but analytically coherent set of cases.

It is important to note that this study does not present empirical evidence of poverty reduction. None of the initiatives examined—whether centralized or decentralized—provide standardized poverty metrics such as income changes, household welfare indicators, or multidimensional poverty scores. As a result, the claims made in this paper relate to the conditions and mechanisms that might enable poverty alleviation rather than to verified developmental impacts. This limitation reflects the broader state of the field: despite widespread expectations that AI-enabled social innovation can support development, there is still a lack of systematic evaluation frameworks and longitudinal datasets that track how these interventions affect poverty-related outcomes over time.

This section interprets the results of both ecosystems through the lens of AI Economics, situates them within broader debates on digital colonialism, decentralization, and social innovation, and reflects on the implications for poverty alleviation strategies in the Global South.

4.1. Established AI4SI Ecosystem: Service Delivery Without Redistribution

The Established AI4SI Ecosystem, captured in Table 1, demonstrates the capacity of AI to deliver tangible benefits in healthcare, agriculture, education, and accessibility. Projects such as DIMAGI (USA) and LifeBank (Nigeria) illustrate how AI-powered tools can extend life-saving interventions and logistical efficiencies into low-resource environments. Similarly, education-focused initiatives like Geekie and RECODE in Brazil or Africa TeenGeeks in Nigeria address pressing human development needs by providing personalized learning and digital literacy programs.

From the standpoint of AI Economics, however, these projects tend to operate within centralized governance models, where data ownership, infrastructure, and decision-making remain concentrated in the hands of corporations, governments, or NGOs. This concentration raises three critical concerns [125]:

• Data Sovereignty and Digital Colonialism

Many of these projects rely on data infrastructures owned or controlled by Global North actors. As Mejias and Couldry [26,126] argue, this can reproduce “data colonialism,” in which value flows outward to centralized data-opolies, while communities remain dependent recipients rather than co-creators of knowledge and value.

2.

Short-term Service Gains vs. Long-term Structural Change

While the initiatives effectively scale interventions, they do not fundamentally alter the structural conditions of inequality [127]. The governance of these ecosystems reinforces dependency on external providers rather than fostering local autonomy.

3.

Scalability without Inclusivity

The Established Ecosystem excels in scalability, but its top-down models often overlook contextual specificity and community participation. For example, Agrimetrics (UK) provides powerful agricultural data analytics, but farmers in rural Africa or Latin America may lack the infrastructural and digital literacy capacities to appropriate and govern these tools.

Thus, although the Established Ecosystem demonstrates that AI can improve service delivery, its economic contribution to poverty alleviation is limited by the lack of redistributive design [128,129]. In the language of Polanyi [29], AI in these contexts remains disembedded from local social relations, subordinated instead to global market logics [130,131,132,133].

4.2. Decentralized Web3 Emerging Ecosystem: Redistributive and Participatory Potential

In contrast, the Decentralized Web3 Emerging Ecosystem (Table 2) seeks to reconfigure the political economy of AI by embedding principles of decentralization, community governance, and data sovereignty [134]. Initiatives such as Grassroots Economics in Kenya, which enables communities to design and manage local currencies, or Fishcoin, which incentivizes sustainable practices through blockchain-based supply chain transparency, directly tackle redistribution by shifting control closer to communities [135].

Several features distinguish this ecosystem from the Established one:

• Community-driven Innovation

Projects like Commons Stack and Gitcoin demonstrate how DAOs and open-source funding mechanisms allow communities to directly shape priorities and distribute resources transparently [136]. This aligns with Stein et al.’s [71,72,73,74,75] advocacy for data cooperatives, which counterbalance data-opolies by redistributing control to individuals and communities.

2.

Embedded Digital Justice

By prioritizing peer-to-peer governance, initiatives such as Giveth in Spain or BanQu in the USA (with Global South applications) challenge the extractive models of centralized charity and supply chains. These models align with Allen et al.’s [78] call for embedding digital justice into the design of decentralized infrastructures.

3.

Alternative Economic Models [137]

The Web3 ecosystem creates new avenues for financial inclusion in regions excluded from traditional banking [128]. Uniswap Grants Program or Powerledger illustrate how decentralized finance and energy trading can bypass traditional gatekeepers, providing marginalized groups with greater autonomy.

Through these practices, the Emerging Ecosystem embodies what Mazzucato would call a mission-oriented approach, explicitly designing systems not just to deliver services but to restructure economic relations in favor of inclusion and sustainability [138].

The distinction between the “Established” AI4SI Ecosystem and the “Emerging” Web3 Ecosystem should be understood as an analytical heuristic rather than a rigid binary. In practice, many social-innovation interventions operate along a continuum that blends centralized infrastructures with participatory governance, distributed data stewardship, or community-led decision-making. Hybrid models are increasingly common in sectors such as health, agriculture, and civic technologies, where central coordination is combined with decentralized or community-based mechanisms. Acknowledging this continuum is essential, as the boundaries between the two ecosystems are porous and evolving rather than fixed.

While the comparative structure of the paper distinguishes two ecosystems for analytical clarity, the argument does not assume a dichotomy. Instead, it highlights how hybrid configurations—integrating centralized capabilities with decentralized governance layers—may offer the most viable pathways for inclusive and sustainable AI-enabled social innovation.

4.3. Contradictions and Risks of Decentralization

Despite its promise, the AR process highlighted that decentralization is not a straightforward solution [122]. The participatory ideal of DAOs and blockchain often collides with realities of exclusion and elite capture [139,140]. As Calzada [122] and Van Kerckhoven & Chohan [83] note, the technological literacy required to meaningfully engage in DAOs risks creating new hierarchies, privileging digitally skilled elites [136,141,142].

Moreover, decentralization can obscure enduring geopolitical asymmetries. The infrastructural backbone of many Web3 projects still depends on Global North venture capital, cloud services, and regulatory contexts. Thus, the illusion of decentralization may mask a re-centralization of power under a different guise, echoing critiques of techno-solutionism and reinforcing dependencies rather than dismantling them [122].

Finally, governance vulnerabilities—ranging from algorithmic biases to security breaches—pose acute risks in fragile institutional contexts. Without careful regulation and participatory safeguards, decentralized infrastructures may exacerbate inequalities rather than alleviate them [143].

Empirical evidence increasingly shows that the participatory ideals of decentralization often collide with socio-technical realities. Van Kerckhoven and Chohan’s analysis of DAO governance demonstrates that decision-making power frequently becomes concentrated among a small group of technically skilled participants, leading to what they call “centralized decentralization” despite the rhetoric of openness [83]. In a Global South context, recent ethnographic work in Kenya shows how blockchain-based initiatives can empower new digitally literate elites who mediate access to technical infrastructures and shape project outcomes, thereby reproducing hierarchy rather than dismantling it. These studies support my argument that without deliberate governance safeguards, literacy investment, and community capacity-building, decentralized architectures risk reinforcing existing inequalities.

The concern that DAOs may privilege technologically savvy actors is consistent with broader evidence on digital self-efficacy and adoption barriers in complex digital systems. Research on mobile banking adoption among Generation Z in the United States finds that digital self-efficacy has a significant negative relationship with adoption when users perceive interfaces as too complex or insufficiently supported [144]. The authors argue that banks must simplify design and provide structured guidance to prevent digitally confident minorities from dominating uptake. This finding is directly relevant to Web3 ecosystems: if meaningful participation in DAOs requires high technical literacy, then users with lower digital self-efficacy are likely to be excluded, reinforcing unequal participation patterns and enabling digitally skilled elites to disproportionately shape governance outcomes. Bringing this evidence into the DAO context adds an empirically grounded mechanism that explains how technical complexity becomes an exclusionary wall.

A number of structural limitations and critiques must also be acknowledged when assessing the potential of decentralized Web3 systems. First, several blockchain infrastructures—particularly those relying on proof-of-work—are associated with high energy consumption and substantial environmental externalities, which are difficult to reconcile with sustainability goals. Second, token-based economic models introduce risks of speculation, price volatility, and market manipulation, which may undermine economic stability and expose vulnerable communities to financial uncertainty. Third, empirical studies consistently show low adoption rates and limited long-term sustainability for many blockchain projects, with a significant proportion collapsing within a short period due to governance failures, inadequate incentives, or insufficient community uptake.

The volatility of the Web3 ecosystem is further illustrated by the collapse of numerous crypto-funded initiatives and decentralized finance (DeFi) experiments, signalling vulnerabilities in economic design, regulatory gaps, and governance immaturity. Finally, substantial technical barriers—including digital literacy requirements, wallet management, cybersecurity knowledge, and access to reliable connectivity—tend to exclude marginalized populations and reproduce existing inequalities. These critiques suggest that while decentralized systems offer conceptual opportunities for redistribution and community governance, their practical implementation faces considerable constraints that must be included in any realistic assessment of their poverty-alleviation potential.

4.4. Comparative Lessons from the Two Ecosystems

Comparing the two ecosystems yields three major insights into the research question [145,146]:

• Centralized Ecosystems offer scalability but risk dependency.

They demonstrate AI’s capacity to improve service provision but rarely redistribute value or empower local ownership. Poverty alleviation here is incremental, not transformative.

2.

Decentralized Ecosystems prioritize redistribution and empowerment.

They provide avenues for local communities to govern data, design currencies, and co-manage resources, aligning more closely with poverty alleviation goals grounded in justice and sovereignty [147].

3.

Neither model alone is sufficient.

Established Ecosystems lack inclusivity, while Emerging Ecosystems face questions of viability and scalability. From the AI Economics perspective [137], a hybrid model that combines the scalability of centralized platforms with the redistributive ethos of decentralized ones may hold the most promise.

While the two ecosystems are contrasted in terms of governance architectures, there is currently no empirical basis for a systematic comparison across key performance dimensions such as cost-effectiveness, scale, direct poverty outcomes, long-term sustainability, or user adoption. Most initiatives in the Established Ecosystem publish operational data or performance metrics, but these metrics are not standardized and rarely relate directly to poverty alleviation. By contrast, initiatives in the Emerging Web3 Ecosystem generally lack longitudinal evaluation, and many remain too early-stage to produce reliable quantitative indicators. As a result, cross-ecosystem comparisons must remain conceptual rather than empirical.

These gaps reflect broader challenges in evaluating social-innovation interventions: cost structures differ widely across sectors, adoption patterns vary with digital literacy, and sustainability depends heavily on regulatory, financial, and infrastructural conditions. Importantly, for decentralized systems, measurable poverty outcomes depend on user participation levels, governance maturity, and token or resource distribution mechanisms—variables that have not yet been systematically assessed.

The comparative framework used in this study highlights governance structures and value flows, but these dynamics operate differently across the Global South. For example, decentralized data-governance models may align well with community-driven innovation networks in parts of East Africa, where mobile-money ecosystems are well established, yet face greater challenges in regions with limited connectivity or fragmented institutional environments. Similarly, AI-enabled social-innovation platforms in Latin America often emerge in contexts with stronger civil-society participation, whereas South and Southeast Asian contexts frequently involve more state-centric digital-governance traditions. A context-sensitive approach is therefore essential for evaluating how AI Economics might enable redistributive outcomes.

4.5. Implications for AI Economics and Poverty Alleviation

From an AI Economics perspective, the redistribution of value, labor, and resources in the Global South depends on three interlocking conditions [148,149]:

• Governance: Decentralized models must embed inclusive decision-making, ensuring communities are not passive recipients but co-governors of digital infrastructures.
• Infrastructure: Investments in connectivity, digital literacy, and local data sovereignty are prerequisites for enabling marginalized communities to benefit.
• Ethics: Algorithmic fairness, privacy, and protection against digital colonialism must be central to any AI-driven intervention [150].

The AR process revealed that when these conditions are absent, both ecosystems risk reinforcing inequities. When present, however, they create pathways for AI to contribute meaningfully to poverty alleviation [151].

4.6. Future Research Directions

Several research trajectories emerge from this analysis:

• Longitudinal Studies—to evaluate the sustainability of Web3-based social innovation over time [141,142]. Do community currencies and DAOs sustain participation, or do they erode under governance capture?
• Hybrid Models—examining cases that blend centralized scalability with decentralized redistribution, testing whether such integrations can resolve current trade-offs.
• Ethnographic Research—following communities as they adopt AI4SI tools, documenting how governance, literacy, and cultural contexts shape outcomes.
• Policy Design—embedding decentralization within global frameworks such as the WEF’s AI4SI or the UN SDGs, ensuring global governance structures account for local needs and sovereignty.

A key priority for future research is the development of longitudinal studies capable of tracking whether redistributive mechanisms introduced by DAOs, community currencies, or decentralized identity systems translate into measurable improvements in income stability, local economic autonomy, or access to essential services. Such research would allow for comparative evaluation between the Established AI4SI Ecosystem and decentralized Web3 models, particularly in relation to risks of elite capture, governance capacity, and long-term sustainability. Without such evidence, claims regarding poverty alleviation must remain cautious and framed as emerging possibilities rather than confirmed impacts.

To enable empirical comparison, future research should develop standardized metrics and longitudinal evaluation frameworks that can assess both centralized and decentralized interventions along several dimensions, including:

• Cost-effectiveness (e.g., cost per beneficiary, operating costs, token-economy sustainability)
• Scale and reach (e.g., geographic expansion, number of active users)
• Poverty outcomes (e.g., income stability, access to essential services, local economic autonomy)
• Sustainability over time (e.g., project longevity, governance resilience, funding continuity)
• User adoption and satisfaction (e.g., user retention, digital self-efficacy, perceived value)

Developing such an evaluation framework is essential for moving beyond conceptual analysis and enabling rigorous comparison of the established and emerging innovation ecosystems.

Future research should adopt region-specific and country-specific comparative designs that examine how AI-enabled social-innovation ecosystems interact with divergent infrastructural, socio-political, and cultural contexts across the Global South. Such work is crucial for assessing whether—and under what conditions—either centralized or decentralized governance models translate into tangible poverty-alleviation outcomes in specific settings.

Future research will need to integrate development indicators and poverty metrics—such as income stability, access to essential services, gendered impacts, employment effects, or changes in local economic autonomy—to evaluate whether AI-enabled social-innovation models translate into concrete improvements in well-being. Longitudinal datasets and mixed-methods impact evaluations will be necessary to assess whether decentralized governance and data-sovereign models produce measurable gains compared to centralized AI4SI interventions.

4.7. Responding to the Research Question

Returning to the central research question—Can AI Economics contribute to poverty alleviation in the Global South?—the findings suggest a conditional yes. AI Economics, when embedded in decentralized, context-sensitive, and ethically governed social innovation systems, can redistribute value, empower communities, and challenge extractive models of data governance. However, if AI continues to be deployed within centralized frameworks that concentrate power and neglect community ownership, it risks entrenching dependency and reinforcing digital colonialism [152].

The comparison of Established and Emerging Ecosystems illustrates that AI’s role in poverty alleviation cannot be understood in technological terms alone. It must be seen as a political economy project, where outcomes depend less on the inherent capabilities of AI and more on the governance architectures, social innovation frameworks, and redistribution mechanisms in which AI is embedded [153].

While platform economics emphasises network effects and data governance concerns institutional control, AI Economics highlights the generation and capture of value through algorithmic prediction and optimisation, thereby offering a distinct lens for analysing redistributive dynamics across innovation ecosystems.

While the decentralized Web3 initiatives presented in Table 2 illustrate new forms of participatory governance, data sovereignty, and community-driven resource allocation, there is currently limited empirical evidence demonstrating direct or sustained poverty alleviation outcomes. Most of these initiatives are still in experimental or early-deployment phases, meaning that longitudinal data, impact metrics, and comparative evaluations remain scarce. As such, their contribution to poverty reduction should be understood as potential rather than established. This limitation is consistent with broader assessments of Web3 ecosystems, where the redistributive promises of blockchain-enabled systems are conceptually significant but remain empirically under-tested.

5. Conclusions

This article set out to examine whether AI Economics—the political economy of value creation, extraction, and redistribution in AI systems—can contribute to poverty alleviation in the Global South. The evidence assembled through the literature review, policy analysis, and AR conducted in the AI4SI International Summer School suggests a conditional but actionable answer [154]. AI’s technological affordances are real and already visible across sectors represented in Table 1—healthcare, education, agriculture, accessibility—yet their development pathways and governance choices determine who benefits and on what terms. In the Established AI4SI Ecosystem, initiatives such as DIMAGI, LifeBank, RECODE, Geekie, Agrimetrics, and others deliver credible, sometimes lifesaving gains: triage and diagnostics at the edge, logistics optimization, and personalized learning tools that respond to acute deficits in capacity and reach. But as our analysis and participants’ reflections repeatedly underscored, these projects typically operate within centralized data and decision architectures that leave the distributional settlement of AI largely intact. The result is service delivery without meaningful redistribution of power over data, models, or value flows, a pattern resonant with the dynamics of “data colonialism” theorized by Mejias and Couldry, in which extraction and commodification of personal and community data reproduce structural asymmetries between the Global North and South [26,126]. Consequently, short-run improvements can coexist with long-run dependency when infrastructures, capabilities, and rules remain controlled elsewhere.

By contrast, the Decentralized Web3 Emerging Ecosystem in Table 2 places redistribution and sovereignty at the center of design. Grassroots Economics’ community currencies in Kenya, Fishcoin’s supply-chain incentives, BanQu’s traceability for marginalized producers, Commons Stack’s and Gitcoin’s DAO-enabled funding of public goods, Giveth’s transparent philanthropy, and energy-sharing experiments like Powerledger and WePower seek to shift the locus of control from platforms and ministries to communities and networks. These efforts operationalize ideas advanced in the scholarship on data cooperatives by experimenting with member-owned governance of digital assets and decision rights [71,72,73,74,75,76,77]. They also echo calls to embed “digital justice” into network design by making participation, transparency, and collective choice constitutive features rather than afterthoughts [78]. De Filippi and colleagues’ account of blockchain as an institutional innovation—redefining trust and authority in “code-mediated” coordination—helps explain why these experiments can matter most in contexts where state capacity and probity are uneven and where centralized intermediaries are either absent or mistrusted [77]. Through the Summer School’s AR dynamics, participants saw how these decentralized forms can open room for local problem definition, member rule-setting, and community capture of data dividends. Yet they also encountered the contradictions mapped by Calzada and by Van Kerckhoven & Chohan: technical literacy thresholds, token-weighted voting pathologies, the risk of governance capture by well-resourced actors, and the nontrivial security and compliance burdens that DAOs and digitally native treasuries impose [83,122]. In other words, decentralization is a necessary but insufficient condition for equitable AI: it repositions the fulcrum of control but does not automatically produce inclusive outcomes.

Set against these empirical patterns, the innovation-studies canon clarifies why the same technology can generate divergent social results [155]. Schumpeter’s account of creative destruction reminds us that innovation is intrinsically redistributive—creating new rents while eroding old ones—and thus political by design rather than by accident [154,156]. Lundvall’s interactive learning perspective shows that capabilities and outcomes are co-produced by firms, states, communities, and knowledge infrastructures; where these innovation system linkages are weak, fragmented, or externally controlled, technological deployment will tend to reproduce rather than repair structural inequalities [155]. Nelson’s enduring question from The Moon and the Ghetto—why societies can land a spacecraft on the moon yet fail to address the basic needs of the poor—applies with force to AI Economics: we have exquisite technical instruments for prediction and optimization, but without institutions that embed justice and redistribution, those instruments will not, by themselves, solve poverty [157]. Pontikakis et al. push this further by foregrounding system dynamics tools for “system innovation,” inviting policymakers to simulate path dependencies, feedbacks, and trade-offs among interventions that blend technological investments with institutional reforms [158]. Read together, these contributions support the core finding of this article: AI’s poverty-reducing potential depends less on algorithmic prowess than on the governance architectures and learning systems within which those algorithms are embedded.

5.1. Comparing Table 1 and Table 2: Established AI4SI and Decentralized Web3 Emerging Ecosystems

The comparative reading of Table 1 and Table 2 therefore yields three integrative conclusions:

First, the Established AI4SI Ecosystem demonstrates that centralized models can scale quickly, standardize quality, and meet urgent needs at low unit cost. These are nontrivial virtues in public health, education, and agriculture. However, absent countervailing institutions, centralized scaling tends to consolidate control over data, models, and standards; communities become data suppliers and service recipients rather than rule-makers, and value circulates within incumbent stacks. This is precisely the scenario flagged by critiques of data colonialism, where digital inclusion without sovereignty morphs into renewed forms of dependency [26,87,126].

Second, the Decentralized Web3 Emerging Ecosystem shows that redistributive design is feasible: member-owned data institutions, DAO-mediated treasuries for public goods, and peer-to-peer marketplaces can change where decisions are taken and where rents accrue. These architectures align with justice-forward approaches to AI governance and can make AI a vehicle for local capability formation rather than mere service targeting [71,72,73,74,75,76,77,78]. Yet their fragility is real: thin local developer bases, volatile token economics, unclear policy regimes, and governance complexity make scaling uneven and sometimes exclusionary [83,122].

Third, hybridization is therefore not a rhetorical compromise but a practical necessity. Hybrid models can pair the logistical and quality-control strengths of established platforms with the ownership, participation, and local value capture of decentralized mechanisms—e.g., centralized clinical decision support that mandates data cooperative membership and community consent for secondary uses; national ed-tech platforms whose personalization layers are co-governed by school-level DAO councils; agricultural analytics that route data royalties through producer cooperatives. Such hybrids speak to Lundvall’s emphasis on institutionally embedded learning and to Pontikakis et al.’s call for system-level design and simulation that foreground feedback loops and distributional impacts [155,158].

The analysis therefore points toward hybridity, rather than strict polarization, as a likely future scenario for AI-enabled social-innovation systems in the Global South.

5.2. Toward a Hybrid Model: Polycentric AI Governance for Social Innovation

The hybridization of centralized and decentralized systems requires a regulatory architecture that acknowledges the distributed nature of innovation while safeguarding ethical and technical standards. Drawing on Ostrom’s notion of polycentric governance, this article proposes a Polycentric AI Governance for Social Innovation (PAIG-SI) framework. In this model, authority and responsibility are shared across multiple, interacting centers:

(i)

Central Nodes—national or regional agencies set baseline ethical, technical, and data-protection standards (e.g., interoperability protocols, AI auditability, and algorithmic accountability).

(ii)

Intermediate Nodes—sectoral data cooperatives or mission-oriented consortia manage domain-specific datasets and coordinate certification of local actors (for example, cooperatives of hospitals or schools).

(iii)

Local Nodes—community-level DAOs or social enterprises implement and adapt AI tools within their socio-economic contexts, ensuring contextual sensitivity and citizen participation.

Regulation and implementation would thus emerge through negotiated coordination among these layers rather than top-down control. For example, a centralized clinical decision-support platform could operate only through membership in a certified health-data cooperative governed by patients, practitioners, and regulators. Similarly, national ed-tech infrastructures could delegate curriculum-level customization to school DAO councils that co-decide algorithmic parameters under shared accountability standards.

This polycentric design allows hybrid systems to retain the scalability and reliability of centralized infrastructures while embedding the participatory and redistributive logics of decentralized ones. It transforms the notion of “hybridization” from an abstract compromise into an actionable governance framework capable of being legislated, audited, and iteratively refined through public–private–community partnerships [9,33,34,35,44,138,155].

While the hybrid model integrates the reliability of centralized infrastructures with the participatory benefits of data cooperatives, it also raises important governance and operational questions that must be addressed for the model to be viable. A centralized clinical decision-support system embedded within a data cooperative would require an administering entity responsible for model validation, quality assurance, and compliance with medical and data governance standards. In practical terms, this administrator could be a public health authority, a regulated non-profit, or a cooperative-appointed technical steward. Regardless of the institutional form, the administrator would need to operate under transparent mandates, regular auditing, and participatory oversight from cooperative members to ensure accountability. In addition, clear procedures would be required to determine how data access is granted, how updates to clinical models are approved, and how conflicts between cooperative members and administrators are resolved.

This clarification also highlights why future research should be attentive to the limitations of both centralized and decentralized systems. Centralized models may struggle with legitimacy and trust, while Web3 ecosystems face risks of fragmentation, uneven participation, and elite capture. Longitudinal research is particularly important for Web3 because initial redistributive benefits—such as broader token distribution or early-stage participatory gains—may diminish over time as technical elites consolidate control, a pattern that cannot be observed in short-term studies. Examining these trajectories empirically will allow researchers to assess whether hybrid arrangements genuinely foster equitable governance or whether decentralization simply shifts, rather than resolves, power asymmetries.

Accordingly, the paper adopts a prospective analytical stance, identifying pathways through which Web3-enabled social-innovation systems could contribute to poverty alleviation, while acknowledging that empirical evidence of realised outcomes remains limited at this stage. Accounting for these limitations reinforces the need for cautious, evidence-based evaluation and for framing decentralized Web3 initiatives as experimental socio-technical configurations rather than ready-made solutions. The absence of standardized empirical metrics therefore remains a key limitation, underscoring the need for future comparative research capable of evaluating redistributive and developmental outcomes across heterogeneous AI-enabled social-innovation models. The contribution of AI Economics to poverty alleviation therefore remains a prospective hypothesis that requires empirical validation through robust measurement frameworks and long-term evaluation designs.

The policy recommendations outlined below derive directly from the empirical insights generated through the Action Research process. Across both online (approximately 150 participants) and in-person activities (approximately 100 participants) conducted during the AI4SI International Summer School, several recurring themes emerged: (i) limited local control over data infrastructures; (ii) concerns about opaque decision-making in centralized AI deployments; (iii) strong interest in participatory governance mechanisms; (iv) uneven digital literacy and technical barriers to meaningful participation; and (v) uncertainty regarding the institutional sustainability of emerging Web3 models. These observations informed the recommendations by identifying governance gaps and priorities expressed by practitioners, policymakers, civil-society actors, and technologists who engaged in the workshops. As such, the recommendations are grounded in the experiential and dialogical evidence produced through the AR cycles, rather than in quantitative evaluation.

Policy and research implications follow directly:

First, governance must be treated as a design parameter, not an ex post compliance layer. Responsible AI frameworks such as those advanced in the AI4SI and PRISM agendas should be expanded to require community governance where data are produced—through data cooperatives or analogous institutions—so that consent, access, and benefit-sharing are structural features rather than discretionary practices [9,33,34,35,44].

Second, infrastructure and capacity are not mere inputs but constitutive elements of equitable AI: connectivity, local compute, open standards, digital literacy, and civic data stewardship shape who can participate and on what terms. Without these, the promise of decentralization devolves into new forms of techno-elitism, as warned by Calzada [122].

Third, ethics must be reinterpreted as distributive justice. Mitigating algorithmic bias and safeguarding privacy are necessary, but insufficient if value extraction remains untouched. Embedding cooperativized data governance [120], local model adaptation rights, and community-negotiated data dividends aligns practice with the digital-justice commitments articulated by Allen et al. and the political diagnoses of Mejias and Couldry [26,78,126].

Fourth, experimentation should be accompanied by system-level learning. Longitudinal, mixed-methods evaluations of DAO-governed programs, community currencies, and cooperative data trusts can establish whether early gains persist, who exits and why, and how governance reforms can counteract capture. Here, system dynamics modeling can help ministries and municipalities simulate equity outcomes before scaling, consistent with Pontikakis et al.’s program for system innovation [158].

The policy recommendations therefore reflect the collective priorities articulated within the AR process and should be understood as grounded in practitioner-informed insights rather than in outcome-based evaluation.

The AR approach used in this study offers a method for pursuing these implications responsibly. By iteratively engaging practitioners, public officials, technologists, and community actors across regions represented in the International Summer School, AR surfaced how abstract governance choices materialize as inclusion or exclusion in situated contexts. Participants observed, for instance, that the same blockchain-based traceability that secures price premia for small producers can also expose them to new compliance risks if buyers, insurers, or regulators weaponize transparency absent protective norms (i.e., market.organic). They also saw that DAO voting mechanics designed for transparency may, in practice, translate technical and financial capital into political power unless thresholds, delegation rules, and deliberative processes are adapted to local literacies. Such observations corroborate the literature on DAOs’ vulnerabilities and the cautions against “decentralization theater” [83,122]. Equally, they demonstrate that design choices—vote weighting, quorum rules, reputational staking, citizen juries, data-commons charters—are levers for redistributive AI, not footnotes [125]. The AR lens thus connects the macro-argument of AI Economics to micro-institutional engineering, where small rule changes can re-route value at scale.

Returning to the research question—can AI Economics contribute to poverty alleviation in the Global South?—the answer is yes, if and only if AI is embedded in social innovation systems that institutionalize redistribution, participation, and sovereignty [154,155,156,157,158]. Centralized programs can and do save lives and expand services; they should not be abandoned. But without reforms that counter extraction and dependency, they risk becoming the latest iteration of what Nelson termed the paradox of technological accomplishment without social resolution [156,157]. Decentralized programs, for their part, open crucial design space for justice-aligned innovation; without attention to capability thresholds and governance complexity; however, they can devolve into enclaves of technical sophistication that exclude those most affected. The route forward is therefore neither naïve techno-optimism nor blanket skepticism but mission-oriented hybridity: pairing the reach and reliability of established infrastructures with the ownership and co-determination enabled by Web3, data cooperatives, and allied institutional forms. This is creative destruction in Schumpeter’s sense [154] but steered toward public purpose and safeguarded by institutions of learning and accountability in Lundvall’s sense [155]. It is also a practical response to Nelson’s “Moon and Ghetto” challenge, aligning cutting-edge problem-solving with structural remedies for inequality [156,157]. And it is a research and policy agenda attuned to system-level feedback, as advocated by Pontikakis et al., capable of anticipating unintended consequences before they harden into new dependencies [158,159].

In sum, AI’s promise for poverty alleviation in the Global South will be realized not by algorithms alone but by the co-design of institutions that decide who sets objectives, who owns and governs data, how benefits are shared, and how risks are borne. The comparative reading of Table 1 and Table 2 shows that we already possess complementary pieces of this puzzle: scalable service platforms and emergent models for community ownership. The task ahead is to stitch them together through policy, law, finance, and civic capacity-building such that digital justice becomes infrastructure, not rhetoric. If the next wave of AI4SI practice and policy can normalize hybrid, community-governed architectures—where data cooperatives negotiate access, DAOs co-budget public digital goods, and national platforms interoperate with local rights and standards—then AI Economics can indeed bend toward poverty alleviation. If not, the Global South risks yet another cycle of technological modernization without emancipation. The choice is institutional, not inevitable; and it is urgent.