Dr Iason Gabriel

Authors: Atoosa Kasirzadeh, Iason Gabriel

Proposes a framework for characterising AI agents along four dimensions – autonomy, efficacy, goal complexity and generality – and uses it to build ‘agentic profiles’ for different kinds of agent, from narrow task-specific assistants to highly autonomous general-purpose systems, as guidance for effective AI governance.

Authors: Iason Gabriel, Geoff Keeling, Arianna Manzini, James Evans

As capable AI agents begin to act in the world on people’s behalf, they raise fresh questions about safety, about human–machine relationships and about social coordination – questions that call for a new ethics of AI agents.

Authors: Hannah Rose Kirk, Iason Gabriel, Chris Summerfield, Bertie Vidgen, Scott A. Hale

As AI becomes more personalised and agentic, people are forming deeper and more persistent relationships with it. Introduces ‘socioaffective alignment’ – how an AI system behaves within the social and psychological ecosystem it co-creates with its user – and argues for AI that supports, rather than exploits, our nature as social and emotional beings.

Authors: Iason Gabriel, Geoff Keeling

Existing answers to the question of what AI should be aligned with – human intentions, or being helpful, honest and harmless – are incomplete and justified in the wrong way. Proposes instead that AI should be aligned with principles that are the product of fair processes, meeting the standard of public justification owed to people in pluralistic societies.

Authors: Iason Gabriel, Arianna Manzini, Geoff Keeling, Lisa Anne Hendricks, Verena Rieser, Hasan Iqbal, Nenad Tomašev, Sofia Ira Ktena, Zachary Kenton, Mikel Rodriguez, Seliem El-Sayed, Sasha Brown, Canfer Akbulut, Andrew Trask, Edward Hughes, A. Stevie Bergman, Renee Marie Shelby, Nahema Marchal, Conor Griffin, Juan C Mateos-Garcia, Laura Weidinger, Winnie Street, Benjamin Lange, Alex Ingerman, Alison Lentz, Reed Enger, Andrew Barakat, Victoria Krakovna, John Oliver Siy, Zeb L. Kurth-Nelson, Amanda McCroskery, Vijay Bolina, Harry Law, Murray Shanahan, Lize Alberts, Borja Balle, Sarah de Haas, Yetunde Ibitoye, Allan Dafoe, Beth Goldberg, Sébastien Krier, Alexander Reese, Sims Witherspoon, Will Hawkins, Maribeth Rauh, Don Wallace, Matija Franklin, Josh A. Goldstein, Joel Lehman, Michael Klenk, Shannon Vallor, Courtney Biles, Meredith Ringel Morris, Helen King, Blaise Agüera y Arcas, William Samuel Isaac, James Manyika

A comprehensive study of the opportunities and the ethical and societal risks posed by advanced AI assistants – agents that plan and act on a user’s behalf. Covers value alignment, well-being, safety, manipulation, anthropomorphism, trust and privacy, and societal-scale effects, and closes with recommendations for researchers, developers, policymakers and the public.

Authors: Laura Weidinger, Kevin R. McKee, Richard Everett, Saffron Huang, Tina Zhu, Martin J. Chadwick, Christopher Summerfield, Iason Gabriel

Applies John Rawls’s Veil of Ignorance to the governance of AI. Across five studies with over 2,500 participants, people choosing principles for an AI assistant from behind the veil consistently preferred a principle that prioritises the worst-off – a choice driven by concerns about fairness rather than risk attitudes or politics.

Authors: Atoosa Kasirzadeh, Iason Gabriel

What does it mean to align conversational agents with human norms and values? Develops a philosophical analysis of linguistic communication between people and conversational agents, identifies ideal norms of conversation, and shows how these norms can guide alignment across different discursive domains.

Authors: Iason Gabriel

Drawing on John Rawls, argues that the basic structure of society is increasingly shaped by AI, so egalitarian norms of justice apply to the technology. AI systems must meet a standard of public justification, support citizens’ rights and promote substantively fair outcomes, with particular attention to the worst-off.

Authors: Laura Weidinger, John Mellor, Maribeth Rauh, Conor Griffin, Jonathan Uesato, Po-Sen Huang, Myra Cheng, Mia Glaese, Borja Balle, Atoosa Kasirzadeh, Zac Kenton, Sasha Brown, Will Hawkins, Tom Stepleton, Courtney Biles, Abeba Birhane, Julia Haas, Laura Rimell, Lisa Anne Hendricks, William Isaac, Sean Legassick, Geoffrey Irving, Iason Gabriel

Maps the landscape of risks associated with large language models across six areas: discrimination and toxicity, information hazards, misinformation, malicious uses, human–computer interaction harms, and automation, access and environmental harms.

Authors: Iason Gabriel

Defends three propositions: the normative and technical sides of AI alignment are interrelated; it matters whether AI is aligned with instructions, intentions, preferences, interests or values; and the central challenge is not to find ‘true’ moral principles for AI, but fair principles that can win reflective endorsement despite widespread moral disagreement.

Authors: Iason Gabriel

Examines the main objections to effective altruism – that it neglects justice, equality, urgency and rights, that it suffers from methodological bias, and that it struggles to achieve systemic change – and weighs the force of each, along with possible responses.