Selected Success Misread as Typical Performance: Why the Take You Saw Is Not the Rate
This conceptual paper introduces Selected Success Misread as Typical Performance, a mechanism in the Selection and Attention series of On Human Modeling. When failed attempts are left out, a viewer can take the one that succeeded as evidence that the performer succeeds easily or reliably. The selected attempt establishes that the feat was possible under those conditions. Judging how typical it was requires a record of the comparable attempts, how many succeeded, and what they cost. In 2006, OK Go released “Here It Goes Again,” a music video in which four band members dance on eight moving treadmills in one continuous shot. Singer Damian Kulash reported 17 takes. A later account put the total at 17 or 18, with the band reaching the end of the routine on two or three. Kulash recalled stumbling during a spin even in the take the band chose. The uninterrupted shot shows that the musicians completed the dance. The production accounts show how often they got to the end. Producer Raymond Chow recalled that Jackie Chan spent 40 days on a shuttlecock scene for the 1982 film Dragon Lord. Chow suggested using editing to finish the sequence, but Chan wanted to capture the jumps and somersaults in one continuous shot. In Chow’s account, he eventually did. The 40 days give a measure of the time spent. The number of attempts remains open, as does what viewers of either film inferred about the performers’ reliability. The paper places the mechanism alongside Robert Rosenthal’s account of the file drawer problem, in which the studies available to readers can differ systematically from those left unpublished. A similar question arises when a team runs an AI assistant several times on the same table and demonstrates a response that copied every figure correctly. A viewer can verify every figure in that response. Estimating how often the assistant gets the whole table right requires the other runs as well.
Authors
- Alexandra Cary
Publication Details
- Journal
- Zenodo (CERN European Organization for Nuclear Research)
- Published
- 2026-11-21
- DOI
- https://doi.org/10.5281/zenodo.24419139
- Primary Topic
- Decision-Making and Behavioral Economics
- Type
- article
- Field-Weighted Citation Impact
- 0.00