By Imogen Rivers
On Friday 7th June, the Institute hosted the ‘AI-Enabled Democracy and Corporate Democratization’ workshop. The workshop features three dynamic and thought-provoking panel sessions bringing together world-leading political philosophers, legal experts and technical developers to discuss some of the most cutting-edge issues in AI governance. The first panel session delves deeper into the insights shared by Professor Joshua Cohen at our Annual Lecture, entitled ‘The Reach of Fairness’. The second session examines how AI may be used to advance democratic ideals. The third session considers how and why corporations may be transformed to better embody democratic ideals.
The Reach of Fairness Revisited
The first panel is chaired by Professor John Tasioulas. Dr Linda Eggert, Early Career Fellow at the University of Oxford Faculty of Philosophy & Institute for Ethics in AI, opens the discussion with probing questions as to the value of algorithmic fairness as a regulative ideal in AI development. In her view, algorithmic fairness has become ‘something of a holy grail’, but there is ‘a nagging concern that it is obsolete’. This is not just due to the well-worn problem that ‘to achieve algorithmic fairness we need to address mathematical impossibility problems’. Rather, the principal problem for Dr Eggert is that whilst social injustice may suffice to generate unfair algorithms, the converse is not true. What point, then, is there in developing fair algorithms if our goal is social justice?
The second panellist is Professor Jeremias Adams-Prassl, Professor of Law at Magdalen College, University of Oxford. Honing in on a central issue of Professor Cohen’s talk, namely the issue of fairness as a response to ‘fortuitousness’, Professor Adams-Prassl offers us ‘a lawyerly response’. He begins by asking how discrimination law has already moved, in the EU and the UK, beyond a focus on group subordination. Would the mechanisms of direct and indirect discrimination speak to the kinds of moral problems which Professor Cohen illuminated at the Annual Lecture? How well do the US law mechanisms of disparate treatment and disparate impact mirror the EU and UK model? Then, Professor Adams-Prassl considers whether corporate structuring can itself be used to disguise responsibility. He concludes with a suggestion that we might need to move beyond tired discussions of how to retain meaningful human control of automated systems—that is, how to keep the human ‘in’ or ‘on’ the loop—into a new conversation about keeping the human ‘above’ the loop.
Our third and final panellist is Professor Josiah Ober, the Markos & Eleni Kounalakis Chair in Honour of Constantine Mitsotakis and Professor of Political Science and Classics at Stanford University. Professor Ober compares the development of the Greek techne of rhetoric with the development of AI. Granted, ‘AI and rhetoric are far from identical kinds of tech’. Still, in a tour-de-force through the history of ideas, our Professor Ober asks whether, vilified by Plato, embraced by Aristotle and shaped by draconian Athenian regulation, could the history of the techne of rhetoric model the future of AI?
Opening the Q&A, Professor Tasioulas considers Lord Bingham’s expansive view of the rule of law, reflecting on how that concept became overwrought. By analogy, he asks, is Professor Cohen packing too much into the concept of fairness: ‘where does fairness stop?’ Why not, indeed, treat the regulative ideal of algorithmic fairness as one value among many? Offering nuanced and insightful responses throughout a wide-ranging Q&A, Professor Cohen calls for both theoria and praxis: ‘There’s a danger in the ethics in AI industry of motivating the work by its practical importance but engaging with the work in a way that is too distant from the practice'.
AI-Enabled Democracy
The second panel session is chaired by Dr Carissa Veliz, Associate Professor at the University of Oxford Faculty of Philosophy & Institute for Ethics in AI. Professor Hélène Landemore, Professor of Political Science and Director of Graduate Studies at Yale University speaks first, proposing an ambitious vision of AI-enabled democracy. Her starting point: representation is not enough and yet we shouldn’t want the whole of society to be structured as a deliberative structure. For this reason, Professor Landemore argues that public deliberation should operate as a structuring norm on our society without strictly regulating it. This prompts the question: how can we harness AI to achieve and augment mass public deliberation? Her provocative discussion concludes with a suggestive dilemma: either we may have to give up on fully effective AI-enabled mass public deliberation, or we may have to give up on deliberative democracy as a theory of political legitimacy.
Our second speaker is Claudia Chwalisz, Founder and CEO of DemocracyNext. Her vision: ‘I see citizens’ assemblies as a crucial component of the kind of change we need.’ So, joining forces with the MIT Center for Constructive Communication, DemocracyNext has launched a two-year project studying how we should approach the use of technology in deliberative assemblies. Claudia takes us through the three strands to their work on AI in public deliberation. Firstly, how can we improve the quality of public deliberation? Secondly, how can we achieve greater transparency? And, thirdly, how can we scale it?
Professor Tasioulas opens the Q&A with a challenge to deliberative democracy as a criterion of legitimacy in Professor Landemore’s work. He contends: 'Deliberative democracy has got a lot going for it, even before we get to legitimacy. So why put so much emphasis on legitimacy?’ His question is avowedly friendly to her work: less focus on legitimacy might expand the impact of her proposals.
Another question from the audience follows: how do we create institutional footholds so that citizens' assemblies can have genuine political power? Professor Landemore's advice, drawing optimism from the French example of a third upcoming citizens' convention, is simple: persist and persist and persist.
Democratising the Corporation?
The third panel session, chaired by Baroness Beeban Kidron, OBE, hosts as speaker Professor Isabelle Ferreras, Visiting Fellow at the Institute and FNRS Professor in Sociology at the University of Louvain. Professor Ferreras defends a radical view of the corporate firm as a political entity involving ongoing dispute over the goals pursued and the means of joint endeavour. But the firm, she argues, is currently under the despotism of the property-owning class who monopolise political rights. Taking a representative model of democracy and drawing on the case study of OpenAI’s Sam Altman, Professor Ferreras sketches a ‘bicameral’ model which re-envisages the corporate firm as one in which workers (i.e., labour investors) elect representatives who then elect the executive body. Looking to the future, she contends that this kind of democratisation of the firm may be crucial not just in ensuring justice for workers as AI transforms corporate governance, but also in advancing the democratic health of wider society, public oversight of corporations, and even responses to the climate crisis.
Our first panellist is Dr Tom Malleson, Associate Professor at King’s University College at Western University, Ontario and co-editor of the volume "Democratizing the Corporation”. Dr Malleson’s emphasis is on the problems which accrue to undemocratic firms. In particular, he argues that such firms exacerbate wider societal injustice, instantiate structural domination and foster economic inequality. Professor Ferreras’ bicameral firm is, he contends, ‘one patch’ for this complex problem.
The second panellist is Dr Caroline Green, an Early Career Research Fellow here at the Institute. She considers how one firm, a retirement village, exemplifies a hierarchical but benevolent structure to address inequality. In this way, Dr Green highlights the danger, when modelling the corporate firm as a political entity, of oversimplification: ‘As political entities, firms are different kinds of governments. Can we have the same expectations for all?’
Our third and final panellist is Dr Ekaterina Hertog, Associate Professor in AI and Society at the Institute. Her specific focus is on the question of diversity in corporate governance. The core contention: ‘Economic bicameralism will not solve everything. But it promises to increase the diversity of representation in corporate governance. And that is something to be celebrated.’
But should we be treating firms as political entities at all? For Professor Tasioulas, this could translate a form of ‘defeatist left-wing politics’. Indeed, he suggests, that way of approaching the problem of capitalist despotism may be part of the reason for seeming state 'toothlessness'.
AI-Enabled Democracy and Corporate Democratisation
After a fruitful and thought-provoking day, a short walk down the Cowley Road takes the Institute and her many guests to dinner at the Central Greek and Mediterranean Restaurant. Professor Tasioulas addresses the AI ethics community with warm words of encouragement and gratitude: ‘We rely on the energy, imagination, and goodwill of many people to keep the show on the road.’