Skip to content

Methodology

How entries in the directory are selected, summarized, and structured.

Sources

Findings and research summaries are drawn from the peer-reviewed and working-paper literature in behavioral economics and decision science (including journals such as Econometrica, the QJE, AER, Science, PNAS, and NBER working papers). We prefer primary sources and link to a freely available full text or working paper where one exists.

How findings are summarized

Each finding is recorded in a consistent, labeled structure so it can be compared and cited:

  • Challenge — the behavioral problem being addressed.
  • Intervention — the specific choice-architecture change tested.
  • Study design — randomized field experiment, online experiment, before–after comparison of participants, retrospective evaluation, and so on. Results are labeled to match: a before–after change is never presented as a control-versus-treatment effect.
  • Outcome — the measured comparison, with honest labels for both conditions, the sample, and where in the paper the numbers appear.
  • Source — the full citation of the study the numbers come from.
  • Biases leveraged — the cognitive mechanisms the intervention exploits.

How biases are written

Definitions are written in original, plain language — the first sentence answers "what is it?" — followed by real-world examples and the design implications. Where the directory contains empirical evidence or research for a bias, it is linked directly.

Quality & corrections

A completeness score tracks whether each entry has its definition, examples, evidence, and citation. Study numbers are checked against the source paper, and DOIs against Crossref. Hypothetical design examples are labeled "illustrative scenario", kept out of every evidence list, and not indexed. If you spot an error in a number or citation, please contact us — we correct in public rather than silently edit.

Corrections log

  • 8 October 2026 — findings database re-verified against the source papers. Corrected figures, samples and labels on all eight findings: the tax-letter result now reports the peer-reviewed trial (Hallsworth et al.) rather than an earlier report's figures; organ donation reports the Johnson & Goldstein online experiment (42% vs 82%); Save More Tomorrow is labeled a before–after change among participants; the Opower result is reported as the 2.0% average reduction versus control; the bail, radiology and algorithm-adjustment entries were corrected to the papers' own numbers, and the bail and radiology studies are no longer listed as nudges. A checkout example that had been presented with invented figures is now labeled an illustrative scenario. Bias tags that did not match the studies were removed, and averaged "lift" figures across unrelated studies were dropped.
  • 8 October 2026 — research summaries. Bibliographic details checked against Crossref and DOIs added; corrected authors and links on two entries (market interactions and trust; trust and betrayal), updated three entries to their published versions, and corrected details in several summaries (Allcott, Camerer et al., Johnson & Mislin, Logg et al., Enke & Zimmermann).