I just published a post-mortem on the now-retracted viral AI‐materials paper by MIT graduate student Aidan Toner-Rodgers.
Toner-Rodgers made big claims on the impact of AI on materials science research productivity and was endorsed by Nobel laureate Daron Acemoglu in the WSJ. 6 months after publication, MIT retracted its support and ArXiv pulled the preprint.
Nantao Li and I were independently investigating Toner-Rodgers’s work weeks before MIT’s disavowal. The article presents the conclusion of our investigation and analyzes the institutional failures that allowed this paper to gain steam and prevented an effective resolution.
Happy to answer any questions or comments.
TLDR:
• The paper contained an obvious breach of informed consent in its survey that should have led the paper to be retracted immediately, instead of doubt about its implausible scope bubbling slowly.
• MIT’s retraction notice failed to disclose any useful information about the misconduct because of federal student protection law (FERPA). This law should be amended in cases of research misconduct.
• Due to the delay from publication to retraction, the paper had time to circulate widely and influence the discussion regarding AI and economic/scientific productivity. It might have provided rhetorical ammunition for the ongoing federal science funding cuts.
nicolasrasmont|9 months ago
Toner-Rodgers made big claims on the impact of AI on materials science research productivity and was endorsed by Nobel laureate Daron Acemoglu in the WSJ. 6 months after publication, MIT retracted its support and ArXiv pulled the preprint.
Nantao Li and I were independently investigating Toner-Rodgers’s work weeks before MIT’s disavowal. The article presents the conclusion of our investigation and analyzes the institutional failures that allowed this paper to gain steam and prevented an effective resolution.
Happy to answer any questions or comments.
TLDR:
• The paper contained an obvious breach of informed consent in its survey that should have led the paper to be retracted immediately, instead of doubt about its implausible scope bubbling slowly.
• MIT’s retraction notice failed to disclose any useful information about the misconduct because of federal student protection law (FERPA). This law should be amended in cases of research misconduct.
• Due to the delay from publication to retraction, the paper had time to circulate widely and influence the discussion regarding AI and economic/scientific productivity. It might have provided rhetorical ammunition for the ongoing federal science funding cuts.