Excited to share our latest work, where we produce sets with both statistical coverage and high decision utility. Applied to dermatological diagnosis, our method yields sets with coherent diagnostic meaning 🏥. More details in the thread 🧵👇

Quantifying AI uncertainty & reasoning about sets of outcomes impt. for high-stakes decisions. Enjoyed collab. w/ Santiago Cortés-Gómez, Bryan Wilder & colleagues on "backing the utility structure of diagnosis" into the learning process: Studies show significant reduction of decision loss.

🚀 <24 hours to go until #EAAMO2024 kicks off! We can't wait to welcome you to UASLP tomorrow, whether you'll be joining us in person or virtually. Conference information can be found at - conference.eaamo.org/conference_inf… Looking forward to 3 days of insightful collaboration! 💡

🧵1/5 In response to Neel Nanda and Cas (Stephen Casper)'s recent work, I published a blog post on how mechanistic interpretability tools can unlock GenAI's potential in high-stakes domains like healthcare. 🚀 See thread below: lesswrong.com/posts/iDyf7aBd…

Can you believe it's already the final day of #EAAMO2024? Today, we're kicking off with a J-PAL panel that brings an innovative focus on how research can drive public policy to create real-world impact, especially when working closely with local and indigenous communities. 🌍

I'm also on bsky now! Same username, Bryan Wilder. If you're getting started there, I'm a fan of Jessica Hullman's starter pack for Algorithmic Decision-Making: bsky.app/starter-pack-s…

Should LLMs be used to review papers? AAAI is piloting LLM-generated reviews this year. I wrote a blog post arguing that using LLMs as reviewers can have bad consequences for science by centralizing judgments about what constitutes good research. bryanwilder.github.io/files/llmrevie…

We show LLMs often fail to reason causally on narratives, relying on heuristics like event order or memorized facts rather than context. Failures can be mitigated by explicitly extracting the underlying causal graph rather than direct textual reasoning. arxiv.org/abs/2410.23884

CMU is hosting a workshop on Human-AI Complementarity for Decision Making this September! Abstract submissions due July 15, travel will be covered for accepted presenters. cmu.edu/ai-sdm/researc…

Project led by Kevin Ren on when LLMs make good predictions and whether we can tell in advance how they'll perform. Lots of phenomena that add texture beyond headline performance numbers. Kevin is off to start a PhD at Cornell now -- can't wait to see what he does next!

Can LLMs reliably improve causal discovery? 🧠✨ Excited to share our new work, MosaCD, showing how LLMs can be geared to provide reliable priors, making causal discovery more accurate and stable. arxiv.org/abs/2509.23570 Great work with Ruiqi Lyu Alistair Turcan Bryan Wilder

🌍 Shape the future of equitable and responsible tech at #EAAMO25! 📅 Nov 5–7, 2025 | Univ. of Pittsburgh ⚡ Early bird registration closes Oct 15. 👉 conference.eaamo.org/registration/