Date: 19 May 2025

Time: 17:00 - 19:00

Venue: Room 104, Northeastern University London, 58 St Katharine's Way, London, E1W 1LP.

Artificial intelligence is seemingly everywhere. And so many bots look, sound, and act like us! From customer service, therapy, and podcast host bots to all-purpose tools like ChatGPT and AI companions that people fall in love with, human-like AI is a growing trend. While this technology has great utility, anthropomorphic AI also presents many risks.

This talk will explore some of the most pressing social and ethical concerns. I’ll argue that the risks are serious enough to justify taking a precautionary approach to design that enhances transparency, promotes meaningful consent, and flips the script on permissionless innovation—the ethos of pushing the envelope with AI until preventable tragedies happen. Simply put, the most responsible way to design AI is to begin from the assumption that bots should, by default, have as few human-like qualities as possible—just enough to perform their core functions. If AI designers wish to add more, they should make a compelling, contextually justified case that doing so won’t present undue risks of the features foreseeably leading to manipulation or exploitation, especially through dark patterns.

Evan Selinger is a Professor of Philosophy at Rochester Institute of Technology. His extensive research and writing focus on the ethical and privacy dimensions of emerging technology. His new co-authored book, Move Slow and Upgrade, will be published by Cambridge University Press (CUP). CUP previously published his co-authored Re-Engineering Humanity and co-edited The Cambridge Handbook of Consumer Privacy. A strong advocate for public philosophy, Selinger is a contributing writer at The Boston Globe.

Ethics of Honest Anthropomorphism: Precautionary Approach to Human-Like AI