Classification of real and falsified narratives in a repeated-measures text corpus
Bibliographic record
Abstract
A great deal of research has been devoted to finding reliable ways of detecting deception. The current deception literature has recognized that human deceptive behavior is highly complex; behavioral differences due to deception (deception cues) are small and probably context dependent. In general, using a repeated-measure design to control variabilities at the individual level is an effective way of amplifying small behavioral effects; however, no study so far has explored the benefit of adding individual-level random-effects in deception detection models. This dissertation focused on a developing field of language-based deception detection research that utilizes natural language processing (NLP; e.g., Fitzpatrick, Bachenko, & Fornaciari, 2015; Heydari, Tavakolia, Salima, & Heydari, 2015). We tested a novel NLP-based deception detection scheme that utilizes multiple language samples from the same individual. A repeated-measures truthful-and-fabricated text corpus (4 truthful and 4 fabricated statements per individual) from 152 individuals was collected. Truth-telling and fabrication scenarios were created using video recordings of real-life negative events as stimuli. Various sets of cues including n-grams, POS tags, and psycholinguistic cues were extracted from the text sample using NLP techniques. Our results showed that mixed-effects variations of popular classification models including logistic regression, decision tree/random forest, and artificial neural network have better cross-context generalizability than their regular fixed-effects counterparts. This research should encourage further development of repeated-measure deception detection schemes and classification models that can fully utilize such a data structure.
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 | 0.007 | 0.055 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.003 | 0.002 |
| Science and technology studies | 0.001 | 0.001 |
| Scholarly communication | 0.001 | 0.001 |
| Open science | 0.001 | 0.001 |
| 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".