Post #2389447
2026-03-27 02:14 UTC
@i_dont_like_ai@mastodon.social @wifsten@mas.to
I like your approach of leaning into points of view that challenge yours. Refreshing atitude to see. More mutually beneficial that #block
I had a good go at finding an AI critical study from MIT in the last 12 months.
I found Demirer et al. (2024), "The Effects of Generative AI on High Skilled Work": MIT Sloan + Microsoft/Princeton/UPenn.
Who found that junior developers were up to 39% more productive and experienced developers were up to 13% output increase. You could spin it negative.
The one a little bit negative was the METR study, cited by Neil Thompson (MIT CSAIL/FutureTech), February 2026: Experienced open-source developers wrote code faster with AI but took 19% longer on the overall task. Developers self-estimated a 20%+ speedup.
Maybe it was an older study.
Or just the Kosmya study, it was everywhere for a while.
#Ai #education #psychology
Replies (1)
-
@i_dont_like_ai@mastodon.social 2026-03-27 11:57
@n_dimension@infosec.exchange @wifsten@mas.to Yes! Underneath is I think right link to study that I have heard of--the article is from July 19, 2025 & hopefully is serious research 😉 This sort of theme was been discussed in 2025 on a front-page-listed article at WSJ, where journalist found that he was loosing French because the computer had been 'overhelpful' with French. In the previous 10-15 years there have been several in-depth studies on the negative effects of overhelpful computers. https://www.hpcwire.com/2025/07/19/ai-is-making-us-dumber-mit-researchers-find/