Tortured Data

“Beware of testing too many hypotheses; the more you torture the data, the more likely they are to confesss, but confession obtained under duress may not be admissible in the court of scientific opinion. ”
—Stephen M. Stigler, “Testing Hypotheses or Fitting Models?” (1987)

That is useful advice for everyone, but even moreso a warning to those seeking to massage cherrypicked data to tell just-so stories. In particular, a few HBDers (human biodversity advocates) can be quite brilliant in their ability to speculate and gather data to support their speculations, while ignoring data that contradicts them. This is seen in the defense of race realism, a popular ideology among HBDers.

Some HBDers and other race realists are so talented at speculating that they come to treat their ideologically-driven interpretations as factual statements of truth, even when they deny this is the case. Just as they deny the consequences of such ideologies being enforced for centuries through social control, political oppression, and economic inequality. A result can be misinterpreted as cause, an easy error to make when evidence for direction of causation is lacking. It leaves the field open to self-serving bias.

When one starts with a hypothesis that one assumes is true, it’s easy to look for evidence to support what one already wants to believe. There are few people in the world who couldn’t offer what they consider evidence in support of their beliefs, no matter how weak and grasping it might appear to others. This is even easier to accomplish when looking for correlations, as anything can be correlated with many other things without ever having to prove a causal connection, and it’s easy to ignore the fact that most correlations are spurious.

None of that matters to the true believer, though. Torturing the data until it confesses is the whole point. As in real world incidents of torture, the validity of the confession is irrelevant.