It was a great experience to be a part of this team, thanks for your amazing leadership Philipp Schoenegger , here you can find our study about human vs. ai persuasion 👇

This is a must-read paper. Everyone fears the use of LLMs to deceive and scam humans. This paper measures how good they are at it. (Spoiler: They're more persuasive than our fellow humans.) thread arxiv.org/abs/2505.09662

👀👀👀 academic research hinting at AI better than humans at sales "Large Language Models Are More Persuasive Than Incentivized Human Persuaders" arxiv.org/abs/2505.09662

AI's are significantly better at persuasion than cash-incentivized humans in a real-time conversational quiz setting, research by @schoeneggerphil et al finds—both when truthful and deceptive (steering towards right vs wrong answers): buff.ly/Lu4TNvF

Proud to be a small part of this fantastic team -- check out our pre-print on #LLM #persuasion at arxiv.org/abs/2505.09662.

I also have another preprint out with Philipp Schoenegger et al. showing similar results on Claude Sonnet 3.5 in interactive quizzes with highly incentivised humans, both in truthful and deceptive persuasion. More on this at: x.com/SchoeneggerPhi…

Outcome-Based Reinforcement Learning to Predict the Future arxiv.org/abs/2505.17989 (news.ycombinator.com/item?id=441068…)

🚨 New preprint from Lightning Rod Labs, in collaboration with Philipp Schoenegger & Luke 🐝 Hewitt 🚨 We trained a compact reasoning model that's state-of-the-art at predicting the future. We massively outperform frontier models at prediction market betting, despite being a fraction

Interesting work here on a 14b LLM for PM forecasting 👇 Check the adapted RL in particular. Nice results on calibration error. This open the door for production tools in the domain imo. Bravo to the team ! 👏 2nd interesting work in a few months on this underrated topic.

💰💰💰Prediction markets are going to get weird. Now we have a smallish open source LLM (14B) that can be trained to predict messy real-world outcomes better than GPT o1. ~thread~ arxiv.org/abs/2505.17989

This study is important because it showed that AI outperforms incentivized humans when persuading people of both true and false claims Persuasion has always been an extremely valuable skill in business, politics, and most other competitive domains arxiv.org/abs/2505.09662

Really enjoyed working on this with Philipp Schoenegger Philip E. Tetlock and Barbara Mellers. We tried ~50 prompt techniques (including AI classics and more theory-motivated ones) on 100 forecasting questions across 6 LLMs. No prompt showed robust improvements! arxiv.org/abs/2506.01578

Coupled with other results like Philipp Schoenegger's other recent paper on the effectiveness of RL for improving forecasting it seems like general prompting might not be the most promising place to push atm. arxiv.org/abs/2505.17989

Prompt engineering has negligible and sometimes negative effects on models' ability to forecast. I feel like this reflects decreasing benefit of prompt engineering as models get more sophisticated, but it could simply mean that we haven't yet discovered a good forecasting prompt.

Some personal news! Next month I will be joining Microsoft AI, working on the economic effects of advanced AI. After an amazing time at LSE, I'm really excited to contribute to this important area of research at Microsoft during such a pivotal moment for AI!