
Christmas lecture by Dr Julia Haas from Google DeepMind on how to evaluate moral competence in large language models.
Dr Julia Haas is coming to The Diamond on Thursday 3 December for a talk on moral reasoning in large language models. It's the second event in the Centre for Machine Intelligence's public lecture series, and it's open to university staff, students and the general public, so you don't need to be connected to the university to get in. The talk is titled A roadmap for evaluating moral competence in large language models.
Haas is a Staff Research Scientist in the Responsibility Unit at Google DeepMind, where she works at the intersection of philosophy and cognitive science. Her research looks at the nature of valuation and its role in theories of the mind, particularly in normative cognition, and the possibility of meaningfully moral artificial intelligence. That's a lot of academic ground to cover, but the talk itself focuses on a specific, practical problem: how to tell the difference between moral competenceāproducing morally appropriate outputs based on morally relevant considerationsāand mere moral performance, where a model might just be mimicking the right answers without any real understanding behind them.
To work through that, she sets out three hurdles that any evaluation of moral competence in large language models has to confront: the facsimile problem, moral multidimensionality and moral pluralism. If you've been following the conversation around AI safety and ethics, these are exactly the kinds of questions that usually get glossed over in the hype, so it's a good chance to hear someone working on them from the inside.
Every week we pick š the 5 best š things to do in Sheffield and send them to your inbox. Just one weekly email, no spam.
By submitting your email, you accept our Privacy Policy
The talk runs from 18:00 to 20:00 on Thursday 3 December at Lecture Theatre 1, The Diamond, The University of Sheffield, 32 Leavygreave Road, Sheffield, S3 7RD. There's a drinks reception from 19:00, and entry is free, though you'll need to register in advance.
Date
Thursday, 3 December 2026
18:00
Price
Free
Recommended Age
+16 years