I recently gave a 15-min talk at #NeurIPS2025 on why "interpretable" AI doesn't automatically lead to better human decisions, and discussed my research on human-AI collaboration.
Watch here: www.youtube.com/watch?v=JTuU...
LLMs are now widely used in social science as stand-ins for humans—assuming they can produce realistic, human-like text
But... can they? We don’t actually know.
In our new study, we develop a Computational Turing Test.
And our findings are striking:
LLMs may be far less human-like than we think.🧵
Preliminary results show that the current framework of "AI" makes ppl less likely to help or seek help from other humans, or to seek to soothe conflict, and that people actively prefer that framework to any others, literally serving to make them more dependent on it.
New research out!🚨
In our new paper, we discuss how generative AI (GenAI) tools like ChatGPT can mediate confirmation bias in health information seeking.
As people turn to these tools for health-related queries, new risks emerge.
🧵👇
nyaspubs.onlinelibrary.wiley.com/doi/10.1111/...
Sure :)
We’ll be presenting @facct on 06.24 at 10:45 AM during the Evaluating Explainable AI session!
Come chat with us. We would love to discuss implications for AI policy, better auditing methods, and next steps for algorithmic fairness research.
#AIFairness #XAI
But if they are indeed used to dispute discrimination claims, we can expect multiple failed cases due to insufficient evidence and many undetected discriminatory decisions.
Current explanation-based auditing is, therefore, fundamentally flawed, and we need additional safeguards.
Despite their unreliability, explanations are suggested as anti-discrimination measures by a number of regulations.
GDPR ✓ Digital Services Act ✓ Algorithmic Accountability Act ✓ GDPD (Brazil) ✓
So why do explanations fail?
1️⃣ They target individuals, while discrimination operates on groups
2️⃣ Users’ causal models are flawed
3️⃣ Users overestimate proxy strength and treat its presence in the explanation as discrimination
4️⃣ Feature-outcome relationships bias user claims
BADLY.
When participants flag discrimination, they are correct ~50% of the time, miss 55% of the discriminatory predictions and keep a 30% FPR.
Additional knowledge (protected attributes, proxy strength) improves the detection to roughly 60% without affecting other measures.
Our setup lets us assign each robot a ground-truth discrimination outcome, which lets us evaluate how well each participant could do under different information regimes.
So, how did they do?
We recruited participants, anchored their beliefs on discrimination, trained them to use explanations, and tested to make sure they got it right.
We then saw how well they could flag unfair predictions based on counterfactual explanations and feature attribution scores.
Participants audit a model to predict if robots sent to Mars will break down. Some are built by “Company X.” Others by “Company S.”
Our model predicts failure based on robot body parts. It can discriminate against Company X by predicting that robots without an antenna fail.
We cannot tell if explanations work or not due to these reasons.
To tackle this challenge, we introduce a synthetic task where we:
- Teach users how to use explanations
- Control their beliefs
- Adapt the world to fit their beliefs
- Control the explanation content
Users may fail to detect discrimination through explanations due to:
- Proxies not being revealed by explanations
- Issues with interpreting explanations
- Wrong assumptions about proxy strength
- Unknown protected class
- Incorrect causal beliefs
Imagine a model that predicts loan approval based on credit history and salary.
Would a rejected female applicant get approved if she somehow applied as a man?
If yes, her prediction was discriminatory.
Fairness requires predictions to stay the same regardless of the protected class.
Right to explanation laws assume explanations help people detect algorithmic discrimination.
But is there any evidence for that?
In our latest work w/ David Danks @berkustun, we show explanations fail to help people, even under optimal conditions.
PDF shorturl.at/yaRua
You're both in!
Denied a loan, an interview, or an insurance claim by machine learning models? You may be entitled to a list of reasons.
In our latest w @anniewernerfelt.bsky.social @berkustun.bsky.social @friedler.net, we show how existing explanation frameworks fail and present an alternative for recourse
Welcome in :)
Oh yeah, welcome to the pack!
Of course!
Actually, I've added you some time ago already so you're good :)
Let's have bioinformatics represented then :) Regarding the clubs, I have not heard of any, might be just a coincidence :D
Added!
Sure Max!
Hey Lucas, consider it done :)
Welcome to the pack :)
Interesting stuff, welcome to the hood!
Of course, welcome in!
