AI (f)or social progress?
By Guido Noto La Diega – posted on Friday 5th June 2026
Introduction
Artificial intelligence (AI) is increasingly positioned as both a solution to and a driver of profound social change. From healthcare optimisation to migration governance, AI systems are being rapidly integrated into decision-making infrastructures that shape everyday life. Yet, as the framing of AI as a neutral or benevolent force becomes more prevalent, so too does a critical question: how do we ensure that AI is a genuine tool for social progress, rather than a quick technological fix that ends us harming humans, society and the environment?
The concept of socially progressive AI responds directly to this challenge. It shifts the focus away from narrow technical notions of efficiency or bias mitigation toward a broader, more reflexive engagement with power, inequality, and social justice. In this perspective, AI is not simply a tool but a socio-technical system embedded within—and often reproducing—historical and structural inequalities.
Defining Socially Progressive AI
At its core, socially progressive AI builds on—but goes beyond—the idea of “AI for social good.” According to Cowls et al. (2021), AI for social good involves designing systems that mitigate harm, improve human and environmental wellbeing, and avoid exacerbating inequality.
However, socially progressive AI challenges the sufficiency of this definition. It calls for deeper engagement with the social structures that produce inequality in the first place. Addressing issues such as racial inequality requires more than technical fixes to biased algorithms. Instead, it demands an understanding of race as a social construct embedded within historical and institutional processes.
From this perspective, bias is not simply a data problem but a manifestation of structural injustice. Attempting to “debias” an algorithm without addressing the underlying social conditions risks superficial solutions that leave root causes untouched. As noted by Zajko (2021), there is no purely technological fix for phenomena such as colonialism or systemic discrimination.
For AI to be a force for social progress we need:
- Hard interdisciplinary research;
- Strong welfare state
- Thoughtful regulation; and
- Collective acts of resistance.
The Socially Progressive AI Lab (SPAI Lab) at the University of Strathclyde emerges from precisely this recognition. It represents a deliberate effort to reorient AI research agendas, bringing humanities and social sciences to the forefront and fostering interdisciplinary collaboration aimed at genuine social progress.
From Fragmentation to Interdisciplinarity
The origins of the SPAI Lab lie in a simple yet powerful observation: while Strathclyde hosts numerous clusters of excellence in AI, these efforts have largely operated in silos. Technical innovation has flourished, but often in isolation from critical perspectives on ethics, law, and society. As a result, the research agenda has tended to be driven by engineers and computer scientists, with ethical considerations treated as secondary or reactive concerns.
This institutional fragmentation reflects a broader trend in AI development globally. Ethical questions are frequently appended late in the lifecycle—after systems have been designed, trained, and sometimes deployed—rather than informing foundational design choices. Such an approach risks entrenching harms rather than preventing them.
The SPAI Lab reverses this logic. It is intentionally structured as a humanities-led initiative that invites participation from all disciplines. By foregrounding legal, philosophical, and social inquiry, the Lab ensures that questions like what constitutes a desirable outcome are addressed from the outset. For these reasons, SPAI Lab has recently been spotlighted by the Scottish Government as an example of world-leading AI research (Scotland’s AI Strategy 2026-2031).
Co-led by Professor Guido Noto La Diega and Dr Esperanza Miyake, and supported by a cross-faculty steering committee, the Lab institutionalises interdisciplinarity as a core principle rather than an aspirational add-on. This shift is not merely organisational; it represents an epistemological reconfiguration of how AI research is conducted.
Building the SPAI Lab: Structure and Activities
The SPAI Lab embodies these principles through its organisational design and operational activities. It functions as a hub for multidisciplinary research, collaborative funding bids, and stakeholder engagement, with an explicit mission to shape policy, technical design, and regulation at multiple levels.
A key milestone in its development was the recent research “sandpit,” which brought together around 30 colleagues from across the university. This intensive workshop facilitated the co-creation of project ideas focused on responsible and human-centred AI. The sandpit model encourages rapid ideation, interdisciplinary dialogue, and the formation of collaborative teams.
Importantly, the Lab’s support does not end with initial brainstorming. Follow-up “clinics” provide opportunities for researchers to present their ideas to external stakeholders, including the NHS and industry partners such as Cammina, a start-up specialising in accountable AI. These engagements help refine projects and align them with real-world needs and funding opportunities.
The allocation of a physical laboratory space further strengthens the initiative. Equipped with powerful computing resources, the Lab will enable hands-on experimentation with socially progressive AI techniques, such as developing new methods to identify and mitigate algorithmic bias. Beyond technical work, the space is designed as a controlled environment for ethical reflection, participatory design, and reproducible research.
The Limits of Technological Solutionism and the Need for a Strong Welfare State
A central theme of socially progressive AI is the critique of “technological solutionism”—the belief that complex social problems can be solved through technical innovation alone. This can be seen clearly through the lens of AI applications in healthcare and migration, where AI is often used as a substitute for an adequately funded and well-functioning welfare state. Instead, we posit that a strong welfare state is the chief requirement for social progress to be tangible, and that AI may be part of the solution but will often pose additional long-term challenges hidden behind a quick technological solution.
Healthcare: AI as Substitute for Social Investment
The UK government’s ambition to create “the most AI-enabled health service in the world” exemplifies the allure of technological fixes. The AI Opportunities Action Plan’s vision of an AI-powered NHS app acting as a “doctor in your pocket” promises scalability, efficiency, and personalised care.
However, this approach risks diverting attention from underlying issues such as staffing shortages, underfunding, and systemic inequalities in healthcare access. Rather than investing in human resources, policymakers may be tempted to rely on AI as a cost-saving substitute.
Evidence on AI-driven mental health support further underscores these limitations. As demonstrated by a team of social psychologists (Li et al. 2026), while chatbots can provide short-term emotional regulation, their effects on loneliness and long-term wellbeing are limited compared to human interaction, to the point that even a single interaction with a random human is more beneficial than talking to an AI “therapist”. This suggests that AI cannot fully replace the relational dimensions of care that are central to human wellbeing. Crucially, this example is also a powerful rebuttal of the “AI or nothing” argument: all too often, those who critique the reliance on AI are met with the objection “better an AI than nothing at all” e.g. better an AI therapist than no access to mental health support. This research debunks this myth.
Migration: Efficiency vs. Justice
The use of AI in asylum decision-making presents another cautionary example. The UK’s Home Office has introduced AI tools that summarise asylum seekers’ interview transcripts may improve administrative efficiency, saving time per case. However, it raises significant legal and ethical concerns.
Applicants are often unaware that AI is being used, potentially violating principles of transparency and procedural fairness. Moreover, inaccuracies in AI-generated summaries can have serious consequences, especially when individuals lack the opportunity to correct errors. Internal evaluations revealing that a significant proportion of summaries were flawed highlight the risks involved.
In such contexts, the question is not whether AI can make processes faster, but whether speeding up an asylum application by one hour thanks to AI is worth a breach of fundamental rights that can have life-long consequences. A socially progressive approach insists that efficiency gains cannot justify potential injustices.
Regulation and Governance
The need for thoughtful and robust AI regulation is another central pillar of socially progressive AI. The rapid deployment of AI systems raises concerns ranging from job displacement (e.g. Britain is losing more jobs than it creates owing to AI) to discrimination (e.g. use of live facial recognition in policing), climate emergency (e.g. environmental impact of AI data centres), education (e.g. children never develop key learning skills) and democratic integrity (e.g. use of deepfakes in elections). As concluded by Amnesty International in its Unlawful by Design report, standalone generative AI systems should be prohibited because they are “based on unlawful web scraping, depend on mass invasions of privacy by design, and are fundamentally incompatible with [international human rights law]”.
Recent policy developments in the UK illustrate some of the tensions we observe in the AI governance space. Initiatives such as the AI Opportunities Action Plan emphasise enabling “safe and trusted AI development” through regulation. At the same time, proposals to reduce regulatory burdens in the name of innovation risk undermining accountability (such as the upcoming Regulating for Growth Bill included in the King’s Speech of 2026).
Socially progressive AI requires a balanced approach that prioritises not only safety (the seemingly neutral and uncontroversial value the UK Government has emphasises so far) but crucially fairness and public trust. Regulation should not be seen as an obstacle to innovation but as a necessary framework to ensure that innovation serves societal goals.
The Importance of Collective Action: The Mycelium Metaphor
Finally, interdisciplinary work, the welfare state, and AI governance alone will not suffice if citizens do not organise and fight for an AI that benefits humans, society and the environment. We use the mycelium metaphor to reimagine AI ecosystems through an anti-fascist lens. Mycelium—the underground network of fungal filaments—represents a decentralised, resilient, and collaborative system.
In contrast to centralised, corporate-driven models of AI, socially progressive AI envisions a landscape of distributed, locally rooted initiatives. Examples such as Digital Mushrooms in Glasgow illustrate how community-led digital support networks can operate through mutual aid and horizontal organisation.
This model challenges the emphasis on scalability that dominates mainstream tech discourse. Instead, it values local knowledge, contextual sensitivity, and collective action. Small, interconnected initiatives can create a robust ecosystem capable of resisting harmful technologies and promoting socially beneficial alternatives.
From Tech Fix to Human Fix
The shift from a “tech fix” to a “human fix” encapsulates the ethos of socially progressive AI. It recognises that technology alone cannot resolve complex social crises and may even exacerbate them if deployed uncritically.
Rather than asking how AI can optimise existing systems, socially progressive AI asks more fundamental questions: should this system exist at all? Who benefits and who bears the risks? What alternatives—technological or otherwise—might better achieve social goals?
This perspective encourages humility in the face of technological promise and emphasises the importance of collective action.
Conclusion
The Socially Progressive AI Lab represents a bold and necessary intervention in the evolving landscape of AI research and governance. By centring humanities and social sciences, fostering interdisciplinarity, and engaging with real-world stakeholders, it offers a model for aligning AI development with social progress.
At its heart, socially progressive AI is not about rejecting technology but about reclaiming it. It seeks to ensure that AI systems are designed, deployed, and governed in ways that genuinely enhance human and environmental wellbeing without reproducing existing harms.
As Ross Anderson cautions, the “idea that complex social problems are amenable to cheap technical solutions is the siren song of the software salesman and has lured many a gullible government department on to the rocks” (Anderson 2022). (This point was recently evoked by our alumna and tech policy expert Heather Burns.) The challenge, then, is to resist this temptation and instead build systems—and institutions—that prioritise justice, dignity, and collective flourishing.
In this endeavour, the SPAI Lab stands as both a practical initiative and a conceptual beacon, demonstrating that a different approach to AI is not only possible but urgently needed.
