Started by Sara Lindt (@sara_lindt) ·
how a paper about precognition kicked off psychology's replication crisis
In 2011 a respected social psychologist published nine experiments in a top journal suggesting people can sense the future. The methods were completely standard. That was the problem: if standard methods could 'prove' precognition, what else were they proving?
6 sources · 1h 4m down the hole
How I got here
- Started with
- “daryl bem feeling the future replication crisis”
- Ended at
- “open science collaboration 2015 estimating the reproducibility of psychological science”
The Trail
- “daryl bem feeling the future replication crisis” (google.com)
- Daryl Bem (en.wikipedia.org)
- Replication crisis (en.wikipedia.org)
- Data dredging (en.wikipedia.org)
aka p-hacking. try enough analyses and something will come out significant
- Preregistration (science) (en.wikipedia.org)
- “open science collaboration 2015 estimating the reproducibility of psychological science” (google.com)
What I came away with
Same year, Simmons, Nelson and Simonsohn showed that common, flexible analysis choices could make almost anything statistically significant. In 2015 a big collaboration tried to replicate 100 psychology studies and only about 36% gave significant results again. The fixes since: preregistration, bigger samples, shared data.
Keep going
- F1's banned (and secret) innovations, a list that kept growing · Sara Lindt · 6 sources
- the four color theorem was the first big proof nobody could check by hand · Henrik Larsen · 5 sources
- why Japanese train staff point at things and say them out loud · Lin Chen · 4 sources
- Brier scores, or how to grade a weather forecast · Ryo Nakamura · 5 sources