Archives des mots-clés : Mengyi Sun

Where there’s a will there’s a way: ChatGPT is used more for science in countries where it is prohibited

Posté le par

Authors : Honglin Bao, Mengyi Sun, Misha Teplitskiy

Regulating AI has emerged as a key societal challenge, but which methods of regulation are effective is unclear. Here, we measure the effectiveness of restricting AI services geographically using the case of ChatGPT and science. OpenAI prohibits access to ChatGPT from several countries including China and Russia.

If the restrictions are effective, there should be minimal use of ChatGPT in prohibited countries. We measured use by developing a classifier based on prior work showing that early versions of ChatGPT overrepresented distinctive words like “delve.”

We trained the classifier on abstracts before and after ChatGPT “polishing” and validated it on held-out abstracts and those where authors self-declared to have used AI, where it substantially outperformed off-the-shelf LLM detectors GPTZero and ZeroGPT. Applying the classifier to preprints from Arxiv, BioRxiv, and MedRxiv reveals that ChatGPT was used in approximately 12.6% of preprints by August 2023 and use was 7.7% higher in countries without legal access.

Crucially, these patterns appeared before the first major legal LLM became widely available in China, the largest restricted-country preprint producer. ChatGPT use was associated with higher views and downloads, but not citations or journal placement.

Overall, restricting ChatGPT geographically has proven ineffective in science and possibly other domains, likely due to widespread workarounds.

URL : https://arxiv.org/abs/2406.11583

Does double-blind peer review reduce bias? Evidence from a top computer science conference

Posté le par

Authors : Mengyi Sun, Jainabou Barry Danfa, Misha Teplitskiy

Peer review is essential for advancing scientific research, but there are long-standing concerns that authors’ prestige or other characteristics can bias reviewers. Double-blind peer review has been proposed as a way to reduce reviewer bias, but the evidence for its effectiveness is limited and mixed.

Here, we examine the effects of double-blind peer review by analyzing the review files of 5,027 papers submitted to a top computer science conference that changed its reviewing format from single- to double-blind in 2018.

First, we find that the scores given to the most prestigious authors significantly decreased after switching to double-blind review. However, because many of these papers were above the threshold for acceptance, the change did not affect paper acceptance significantly.

Second, the inter-reviewer disagreement increased significantly in the double-blind format.

Third, papers rejected in the single-blind format are cited more than those rejected under double-blind, suggesting that double-blind review better excludes poorer quality papers.

Lastly, an apparently unrelated change in the rating scale from 10 to 4 points likely reduced prestige bias significantly such that papers’ acceptance was affected.

These results support the effectiveness of double-blind review in reducing biases, while opening new research directions on the impact of peer-review formats.

URL : Does double-blind peer review reduce bias? Evidence from a top computer science conference

DOI : https://doi.org/10.1002/asi.24582