Slang & Culture
P-hacking
Running many analyses, or adjusting data and methods, until a result crosses the threshold for statistical significance (usually p < 0.05), then reporting only that result. It makes chance findings look real and is a known cause of research results that fail to replicate, including in AI benchmark claims.
Origin
The term was popularized by Uri Simonsohn, Leif Nelson and Joseph Simmons around 2011–2012, following their paper "False-Positive Psychology."