Joshua Gans

We expect higher safety standards from automated agents than from humans. @joshgans.bsky.social argues this can be rational, effectively because, while humans can be motivated to provide costly help, driven by (anticipated) guilt if they didn’t, machines feel no guilt:

My thoughts on OpenAI's Big Maths Day. joshuagans.substack.com/p/openai-dro...

The FTC argues Amazon did not declare its auction rules. This paper shows when that matters, from @joshgans.bsky.social www.nber.org/papers/w35841

Studying how advertising claims direct consumers to different customer experiences, and how prices change the available evidence, from @joshgans.bsky.social www.nber.org/papers/w35770

Americans saved $1.5 trillion not buying Trump steaks

The Mathematicians are struggling, My response to the Fields Medalists' letter joshuagans.substack.com/p/the-mathem...

joshuagans.substack.com/p/you-dont-m...

Assessing the role of AI-generated assessments of manuscripts in improving the quality of peer review, from @joshgans.bsky.social www.nber.org/papers/w35688

The Hugging Face incident is a 5-alarm fire. joshuagans.substack.com/p/its-worse?...

Come join our group at Rotman. Applications open. jobs.utoronto.ca/job/Toronto-...

Restricting frontier AI models for exclusive evaluation prior to public release changes the cybersecurity game, from @joshgans.bsky.social www.nber.org/papers/w35586

Not quite the SciFi scenario joshuagans.substack.com/p/not-quite-...

The cybersecurity AI risk is far worse than we thought. joshuagans.substack.com/p/its-time-f...

Studying jobs in which workers value both producing useful output and knowing that output depends on their own contribution, from @joshgans.bsky.social www.nber.org/papers/w35559

Once the price of passive equity stakes to the government and a firm's policy exposure are determined in equilibrium, the owners' consent reveals nothing about whether they benefit, from @joshgans.bsky.social www.nber.org/papers/w35518

What makes AI economically valuable? AI is used by humans to make decisions. The optimal way to train or benchmark AI depends on that joint use, in a formal decision-theoretic way, from Kevin A. Bryan and @joshgans.bsky.social www.nber.org/papers/w35490

If you win the FIFA peace prize you are allowed a mulligan

Trying to purify a metric to reducing gaming in markets can backfire when a metric feeds into prices but not when it feeds into rank/position, from @joshgans.bsky.social and Scott Duke Kominers www.nber.org/papers/w35363

In today's Tea for Teaching podcast, @joshgans.bsky.social joins us to discuss the impact of generative AI on learning, labor markets, and All-Day TA, an app designed to enhance student learning and to provide feedback to professors. teaforteaching.podbean.com/e/all-day-ta/ @cyberthread.bsky.social

In today's Tea for Teaching podcast, @joshgans.bsky.social joins us to discuss the impact of generative BO on learning, labor markets, and All-Day TA, an app designed to enhance student learning and to provide feedback to professors. teaforteaching.podbean.com/e/all-day-ta/ @cyberthread.bsky.social

I watched this talk 17 years ago and then couldn't find it again but it just popped up. It is the best talk I have ever heard on marketing and beyond. Do yourself a favour and watch it. You won't regret it. www.ted.com/talks/rory_s...

Economists have studied how to persuade decision-makers; considering who transmits information without persuasion, from @joshgans.bsky.social www.nber.org/papers/w35312

If workers enjoy tasks, automation can both increase work satisfaction while reducing pay, from @joshgans.bsky.social www.nber.org/papers/w35309

Young @tonymartin.bsky.social graduates from Harvard. www.youtube.com/watch?v=Bz_i...

Building an MCP to research papers. justinross.substack.com/p/an-ai-inte... Yet another opportunity for @RefineDotInk

Arguing that the assumption that empirical measures of AI’s wage effect typically hold fixed the bundle of activities a worker is paid for at its pre-AI shape hides much of the action, from @joshgans.bsky.social www.nber.org/papers/w35211

It is often optimal to choose two contrasting experiments than more other experiments to determine entrepreneurial strategy, from @joshgans.bsky.social and Luca Gius www.nber.org/papers/w35207

Interesting new paper on precisely when and from whom the innovations that led to the current commercial space revolution came from. www.nber.org/papers/w3521...

University grades permit informative comparisons across courses only when transcripts recover comparable student signals through course-adjusted “eigengrades,”. Exact difficulty adjustment removes incentives to choose easier courses, from @joshgans.bsky.social and Kominers www.nber.org/papers/w35183

Developing a dynamic model in which a founder chooses between stealthy and public experiments while potential entrants infer from both actions and outcomes, from @joshgans.bsky.social www.nber.org/papers/w35172

I read this paper yesterday. Very interesting including the notion that the very existence of AI might actually encourage more learning even if it isn't used!

The most favorable environment for entrepreneurial entry is neither low or high variance but somewhere in between, from @joshgans.bsky.social www.nber.org/papers/w35091

Books in hand.

Fantastic panel on the role of agentic AI in research, hosted by @strategicmgmt.bsky.social and Sukhun Kang, with @mattbeane.bsky.social, Gwendolyn Lee, @cgartenberg99.bsky.social, @joshgans.bsky.social. Image from Claudine Gartenberg. It's a new world (with lots of unanswered questions).

Cc @alroth.bsky.social

Two thoughtful essays about the impact of LLMs on graduate education: ergosphere.blog/posts/the-ma... economics.mit.edu/sites/defaul... From personal experience I think the self-control problem mentioned in the first essay is very real.

This would seem to be required for basic preparation. Good grief.

Interesting new book by Alex Krausse, The Engine of Scientific Discovery. Available here doi.org/10.1093/oso/...

How can we tell good AI from bad? as.ft.com/r/10e94b4b-d...

after a night of russian missile attacks on Ukraine, I decided my saturday needed something stupid and fun. so I built a random econ paper title generator. current favorite: Craft Beer Variety as an Instrument for Public Trust in Economists small joys. dariia-m.github.io/econ-title-g... #Econsky

This is worth 5 minutes of your time. www.terrygodier.com/the-last-qui...

New short fun paper alert on data pricing. authors.elsevier.com/sd/article/S...

I don't think we will have a Journopoclypse open.substack.com/pub/joshuaga...

Don’t change the regime and just take the oil?

U.S. Still Enjoying Small But Loyal Following

I really get the feeling that Anthropic should have been charging more than $200m for what they provide.

WTF is pro-Worker AI? open.substack.com/pub/joshuaga...

Prediction, Decision-MIf artificial intelligence is best understood as a technology of prediction, how does that reshape who makes decisions, who captures value, and who sets the rules of modern markets? @joshgans.bsky.social @utoronto.ca @econuoft.bsky.social discusses: faculti.net/the-microeco...

Prediction, Decision-MIf artificial intelligence is best understood as a technology of prediction, how does that reshape who makes decisions, who captures value, and who sets the rules of modern markets? @joshgans.bsky.social @utoronto.ca @econuoft.bsky.social discusses: faculti.net/the-microeco...