Interestingly, when we run the same exercise with LLMs choosing the pricing algorithms, they are even more collusive than humans when left to design the algorithms themselves. Greater automation could lead to even worse outcomes.
More details in Adrian's post.
On my way to Dublin for the EEA meeting! Hope to meet many of you.
Quick side note: Imagine the rail line between Galway and Dublin with double tracks and reliable mobile coverage through the Midlands: paradise!
And that's right, but in the last ~15 minutes, he presented some slides on the paper by @EmilioC_ and co-authors which was later published in the AER on RL-algorithms in pricing environments.
My single-authored paper "Algorithmic and Human Collusion" has been accepted for publication at the Economic Journal.
More than happy about this outcome and the process of getting there.
I won't summarize the paper yet again on here, but a small anecdote about the paper🧵
Frisch gelandet in Düsseldorf für die @DICEHHU competition policy conference.
Erstmal ein Taxi Teller!
📉 Result 2 (Prices): Pro-competitive effects of AI!
We thought the collusive AI would raise prices.
However, by the end, prices are significantly lower in both AI treatments than in the human-only BASELINE.
📈 Result 1 (Adoption): Participants delegate a lot.
But control matters. Adoption is significantly higher when participants can override the AI (RECOMMENDATION).
📣 New paper! "Delegate Pricing Decisions to an Algorithm? Experimental Evidence" with @hans_normann, Nina Rulié, & @ostypa
We study what happens when humans delegate to a collusive algorithm, and AI pricing adoption becomes a strategic choice!
arXiv: arxiv.org/abs/2510.27636












