Friends: We have to stop running experiments where the outcome is a difference score. It is strictly worse in terms of statistical power. declaredesign.org/blog/posts/u... Control for the pre-treatment outcome instead! (example)

Show Post

@igi.bsky.social @gordpennycook.bsky.social @smfleming.bsky.social tomcostello.bsky.social

Show Post

Episode 2 of On Wisdom's five-part special series is live! This week features a discussion on Metacognition and Judgment with hosts Igor Grossmann and Charles Cassidy. They are joined by guests Gord Pennycook, Steve Fleming, and Thomas Costello. 🎧Listen here: www.onwisdompodcast.com/70

Show Post

🧵[3/3] persuade. That is the intervention being tested! We aren't checking if an LLM accidentally deradicalizes someone while generating a recipe for soup. We are testing a targeted tool.

Show Post

I am an AI pessimist in many respects, but the experimental evidence on this (thanks @tomcostello.bsky.social) is fairly solid. The challenge, of course, is that interaction with LLMs needs to be structured properly — a big issue given how undisciplined use of these tools is often promoted.

Show Post