Patrick S. Forscher
@psforscher.bsky.social
Senior Behavioral Scientist at Busara, a non-profit that does behavioral science to advance development. Strong interest in applied meta-research. https://patrickforscher.com/
I mean I think there is a reason the replication crisis was so bad in social psychology and that reason very likely has something to do with incentives
osf.io/6hevp
osf.io/6hevp
OSF
osf.io
January 15, 2026 at 7:33 PM
I mean I think there is a reason the replication crisis was so bad in social psychology and that reason very likely has something to do with incentives
osf.io/6hevp
osf.io/6hevp
Can’t recommend it enough
January 13, 2026 at 7:09 AM
Can’t recommend it enough
Reposted by Patrick S. Forscher
We @chineseopensci.bsky.social wrote a brief Chinese introduction for this initiative: mp.weixin.qq.com/s/j8oTyoGtJ8...
开放数据 | WARN-D机器学习比赛启动，寻找抑郁症预测最佳模型
WARN-D 项目开放抑郁症研究数据，诚邀各界团队参与机器学习比赛，寻找最具预测能力和临床价值的模型
mp.weixin.qq.com
January 9, 2026 at 5:21 AM
We @chineseopensci.bsky.social wrote a brief Chinese introduction for this initiative: mp.weixin.qq.com/s/j8oTyoGtJ8...
Reposted by Patrick S. Forscher
Explainers & tutorials are a great way to criticize current practices with a positive twist. Tailoring them to a specific substantive (!) subfield can greatly increase uptake. Forget about novelty; if some statistican said sth in the 70s but no one was around to hear it, say it again.>
January 8, 2026 at 7:28 AM
Explainers & tutorials are a great way to criticize current practices with a positive twist. Tailoring them to a specific substantive (!) subfield can greatly increase uptake. Forget about novelty; if some statistican said sth in the 70s but no one was around to hear it, say it again.>
Reposted by Patrick S. Forscher
I cannot readily imagine a relative advantage of sampling survey/experiment data from an LLM to sampling known models/distributions if we cannot meaningfully define a population LLMs sample or formalize the sampling scheme. Idk what we can learn abt a model/method without clarity abt what we put in.
December 16, 2025 at 9:54 PM
I cannot readily imagine a relative advantage of sampling survey/experiment data from an LLM to sampling known models/distributions if we cannot meaningfully define a population LLMs sample or formalize the sampling scheme. Idk what we can learn abt a model/method without clarity abt what we put in.