Data Ethics Group

Recent updates

DEG Theme for 2022-23: AI, Power, and the Public Interest: Who’s in control?

Despite the current swell of appeals to data science and “AI for the social good’ in the academic literature, the sociotechnical reality of contemporary data scientific practice seems to be telling a very different story. Present-day AI research and development ecosystems are contested fields composed of complex and contending interests, asymmetrical power relations, and prohibitive entry costs.

Notwithstanding their projections of charitable rhetoric, large tech companies are increasingly engaged in a “commercially-driven production of the social good” that, many have argued, produce outcomes which run counter to the public interest. The corresponding rise of phenomena like “data colonialism” and “philanthro-capitalism” are complicating “AI for the social good” narratives. They are heralding the coalescence of high-entry-cost digital innovation ecosystems, multinational corporate business models, and marketing strategies with humanitarian data work, digital development schemes, and market forces.

The result is the proliferation of privately controlled forms of extractionist data work, that are, in fact, often at cross-purposes with inclusive, equitable, and societally beneficial innovation. Trends such as these appear to be signalling “a profound rebalancing of power and governance in the domain of social life, privileging corporations with large-scale data power and making states (and other commercial and civil society actors) dependent on those corporations.” Similarly, incipient forms of data governmentality and commodification ever more infiltrate academic venues and research environments.

In these, large tech companies’ proprietary data sets, seemingly unlimited financial resources, and massive computing power are, in effect, ‘de-democratising’ AI and data science. Several processes and dynamics hasten this. They include the control that corporations possess over access to data and compute resources, their command over labour power through the university-corporate hybridisation of ‘dual-affiliation’ career trajectories, and their manipulation of the terms of open research to protect their own rentiership claims to monopolistic control over intellectual property and infrastructural assets.

The constellations of asymmetrical power relations, private corporate interests, and high entry costs for data innovation seem to establish the terms of engagement for accessing critical digital infrastructure that should otherwise be tightly bound up with the pursuit of the public good. The theme for our DEG dialogues this year, “AI, Power, and the Public Interest: Who’s in control?,” is framed around the difficult questions raised by these trends.

Questions of concern include: Is the current universe of data scientific innovation—that is, the global assemblage of data infrastructures, compute infrastructures, algorithmic infrastructures, funding schemes, and research and delivery capabilities and resources—equipped to actualise responsible and sustainable data work that serves the public good? Or do these infrastructural, resourcing, and human factors operate, in fact, to curtail and elide possibilities for that actualisation?

Is the human and biospheric interest in the realisation of the global public good being fostered by the current national and global configuration of power relations, sociotechnical affordances, platform ecosystems, and infrastructural networks that characterise contemporary data scientific research and innovation environments? How should terms such as “public good,” “social good”, and “public interest” be construed and reinterpreted in the light of the factors that purport to promote their realisation, or are seen as inhibiting them?

What role should states and governments play in the provision of, and the regulation and governance of, critical digital infrastructure? How can, and should, effective technology governance (including laws, regulatory regimes, and other policy instruments) work in the complicated, multivalent, and transnational innovation ecosystems and platform political economy that is dominated by big tech?

Past events

Towards Effective Supervisory Oversight? Exploring UK Regulatory Enforcement of Data Protection and Electronic Privacy Rights in the context of the Government’s Statutory Reform Plans, 28 September 2023.

Although the (UK) GDPR mandates strong enforcement and a prioritisation of this by the regulator including through the handling of data subject complaints, severe limitations exist in practice. Indeed, in 2021-22 the Information Commissioner’s Office (ICO) did not serve a single GDPR enforcement notice, secured no criminal convictions and issued only 4 GDPR fines totalling just £633k. The Tribunal has removed any substantive bite to the individual order to progress complaints remedy and the DCMS Committee has failed to provide effective holistic scrutiny. In terms of the Government’s reform plans, there may be a case for some of the legislative reforms now proposed including reconstituting the ICO as a corporate board and increasing transparency. However, others risk providing a de jure entrenchment of the ICO’s positioning away from being a comprehensive upholder of core data protection rights. None directly address the serious challenges present here. 

This session will explore the current lay of the land and explore what might lead to a more effective supervisory enforcement.  It will be argued that, at least in public interest cases, the order to progress complaints should police the appropriateness of the ICO’s substantive as well as procedural response and not-for-profit representative complaints should be permitted even without the mandate of data subjects in order to encourage well-argued, strategically important cases. In addition, and at least as importantly, it will be argued that the Equality and Human Rights Commission should be obliged to periodically provide holistic scrutiny of the ICO’s enforcement track-record from a human rights perspective within which data protection rights must ultimately sit.

Our Public Dialogues have previously brought together a range of experts to engage in an open format discussion on how their work addresses the DEG theme. Previous events include: 

• 11 May, 2023: The Data Ethics Group convened for an ‘Open Dialogue: Foundation Models and the Public Interest’. There is heightened anxiety about potential disruptions to markets, governments and civil society resulting from rapid developments in transformer architecture and roll-out of models trained on a burgeoning volume of data. While foundation models are trained on large corpora of data to be task-agnostic, their proliferation and interaction include are likely to exacerbate existing bias, discrimination and mis/disinformation which can augment structural inequities and power imbalances at both a domestic and global level. Further, the limitations of existing regulations, threats of model hallucination and negative environmental externalities, among others, call for a close inquiry into these contemporary developments and the future landscape of AI. 
   
• 26 January, 2023: Professor Frances Griffiths presented and chaired a discussion on the topic of ‘AI, Inequality, and Power Relations in Global Healthcare.’ Across the globe, there are inequalities in healthcare provisions, and within national populations, inequalities of access persist. Additionally, power asymmetries between patients and healthcare workers manifest within the domain of healthcare. In this dialogue, we reflected and discussed on the social and ethical issues arising from the growing prevalence of the use of AI to support clinical decision-making and AI in medical devices within the larger healthcare ecosystem. Dr Michael Katell and Smera Jayadeva also presented their research on Equity in Medical Devices which focuses on the convergence of the social determinants of health and AI within the healthcare landscape of the UK.
   
• 9 May, 2022: Dr James Wright presented on the research from the Privacy, Agency and Trust in Human-AI Ecosystems (PATH-AI) project which was followed by a discussion chaired by Dr David Leslie and Professor Charles Raab. PATH-AI is a UK-Japan collaboration that will develop and pilot a methodology for inter-cultural co-design of a culture- and human-centric framework for more ethical and equitable human-AI ecosystems. This project is a cooperation between The Alan Turing Institute, the University of Edinburgh, and RIKEN – Japan’s largest comprehensive research institution, and is supported by UK Research and Innovation (UKRI) and Japan Science and Technology Agency (JST).