Detecting deception: adversarial problem solving in a low base‐rate world
Bibliographic record
Abstract
Abstract The work presented here investigates the process by which one group of individuals solves the problem of detecting deceptions created by other agents. A field experiment was conducted in which twenty‐four auditors (partners in international public accounting firms) were asked to review four cases describing real companies that, unknown to the auditors, had perpetrated financial frauds. While many of the auditors failed to detect the manipulations in the cases, a small number of auditors were consistently successful. Since the detection of frauds occurs infrequently in the work of a given auditor, we explain success by the application of powerful heuristics gained from experience with deceptions in everyday life. These heuristics implement a variation of Dennett's intentional stance strategy, which is based on interpreting detected inconsistencies in the light of the Deceiver's (i.e., management's) goals and possible actions. We explain failure to detect deception by means of perturbations (bugs) in the domain knowledge of accounting needed to apply these heuristics to the specific context of financial statement fraud. We test our theory by showing that a computational model of fraud detection that employs the proposed heuristics successfully detects frauds in the cases given to the auditors. We then modify the model by introducing perturbations based on the errors made by each of the auditors in the four cases. The resulting models account for 84 of the 96 observations (i.e., 24 auditors × four cases) in our data.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.012 | 0.076 |
| Meta-epidemiology (narrow) | 0.001 | 0.001 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.001 | 0.001 |
| Science and technology studies | 0.002 | 0.005 |
| Scholarly communication | 0.004 | 0.006 |
| Open science | 0.003 | 0.004 |
| Research integrity | 0.003 | 0.003 |
| Insufficient payload (model declined to judge) | 0.002 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".