Discerning truth from deception: Human judgments and automation efforts
Bibliographic record
Abstract
Recent improvements in effectiveness and accuracy of the emerging field of automated deception detection and the associated potential of language technologies have triggered increased interest in mass media and general public. Computational tools capable of alerting users to potentially deceptive content in computer–mediated messages are invaluable for supporting undisrupted, computer–mediated communication and information practices, credibility assessment and decision–making. The goal of this ongoing research is to inform creation of such automated capabilities. In this study we elicit a sample of 90 computer–mediated personal stories with varying levels of deception. Each story has 10 associated human deception level judgments, confidence scores, and explanations. In total, 990 unique respondents participated in the study. Three approaches are taken to the data analysis of the sample: human judges, linguistic detection cues, and machine learning. Comparable to previous research results, human judgments achieve 50–63 percent success rates, depending on what is considered deceptive. Actual deception levels negatively correlate with their confident judgment as being deceptive (r = -0.35, df = 88, ρ = 0.008). The highest-performing machine learning algorithms reach 65 percent accuracy. Linguistic cues are extracted, calculated, and modeled with logistic regression, but are found not to be significant predictors of deception level, confidence score, or an authors’ ability to fool a reader. We address the associated challenges with error analysis. The respondents’ stories and explanations are manually content–analyzed and result in a faceted deception classification (theme, centrality, realism, essence, self–distancing) and a stated perceived cue typology. Deception detection remains novel, challenging, and important in natural language processing, machine learning, and the broader library information science and technology community.
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 | 0.028 | 0.224 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.000 |
| Bibliometrics | 0.004 | 0.001 |
| Science and technology studies | 0.002 | 0.004 |
| Scholarly communication | 0.006 | 0.004 |
| Open science | 0.001 | 0.003 |
| Research integrity | 0.001 | 0.001 |
| Insufficient payload (model declined to judge) | 0.002 | 0.001 |
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".