Cherry-picking
Incomplete evidence, Cueillette de cerises
Presenting only the data that support a conclusion, passing over those, sometimes the majority, that contradict it.
The essentials
Argument diagram
From data D, keep only the subset favourable to P.
Conclude P.
Why the argument is invalid
Each piece of data presented may be accurate: it is the selection that misleads. In a large enough data set, one almost always finds items compatible with any conclusion.
In practice
Example
A manufacturer cites the three studies favourable to its product and passes over the twelve that find no effect.
Famous case
Those who deny global warming have long claimed that the climate has not warmed “since 1998”. That year was exceptionally hot because of a strong El Niño episode: taking it as a starting point hid the long-term trend.
Defending yourself
How to respond
Ask for the full data set, the full period, and how the selection was made. Prefer meta-analyses to isolated studies.
Don’t confuse
Not to be confused with: Hasty generalisation
Cherry-picking discards the unfavourable data; hasty generalisation concludes from too little data.
Going further
Source
Publication bias, whereby positive results are published more often than null results, is an institutional form of it: Robert Rosenthal called it “the file drawer problem” (Psychological Bulletin, 1979).
Practise
In the app, each entry offers ways to practise recognising the concept. No sign-up: your progress stays on your device.