🧪 This platform is in early beta. Features may change and you might encounter bugs. We appreciate your patience!
regression_to_mean
The regression to the mean fallacy occurs when people interpret a natural statistical phenomenon as a causal effect. Extreme values on any measurement tend to be followed by less extreme values simply due to random variation, not because of any intervention. This leads people to falsely attribute the return to average to whatever action was taken between measurements.
A sports team has its worst season in a decade and hires a new coach. The next season, performance improves to near-average. Fans credit the new coach, but statistically, the team was likely to regress toward its mean performance regardless of any coaching change.
A sales manager notices her worst-performing rep had a terrible quarter, so she delivers a harsh performance review. The next quarter, his numbers improve to near-average. She concludes that 'tough love works,' not recognizing that an unusually bad quarter was statistically likely to be followed by a more typical one regardless.
A school introduces an intensive tutoring program specifically for students who scored in the bottom 10% on a standardized test. The following year, those students score notably higher on average. The administration celebrates the program's success, overlooking that students who score at an extreme low are statistically likely to score closer to average on a subsequent test.
Binary (yes/no) questions an LLM must answer to identify this aspect:
Was a measurement or observation taken at an extreme value?
Type: binaryDid a subsequent measurement move toward the average?
Type: binaryIs the return to average attributed to a specific intervention rather than natural fluctuation?
Type: binaryThe regression to the mean fallacy occurs when people interpret a natural statistical phenomenon as a causal effect. Extreme values on any measurement tend to be followed by less extreme values simply due to random variation, not because of any intervention. This leads people to falsely attribute the return to average to whatever action was taken between measurements.
Humans instinctively seek causal explanations for every observed change. When an intervention coincides with natural regression, it is almost impossible psychologically to separate the two effects.
Use control groups that receive no intervention to see if improvement occurs anyway. Compare against the long-term average, not just the most recent extreme measurement.
This fallacy frequently appears in evaluating medical treatments (patients seek help when symptoms are worst), educational interventions (tested after poor performance), and performance management (punishing bad performance appears to work because it naturally improves).
Assuming cause-and-effect because events are correlated or sequential (post hoc ergo propter hoc).
Making definitive linear predictions about complex non-linear systems.
Extending conclusions beyond the range of observed data without justification.
Believing that small samples accurately represent the underlying population distribution.
The tendency to overestimate the accuracy of one's judgments, especially when available information is internally consistent, even if the information is limited or unreliable.
Use these tools to detect, analyze, or train this aspect.