Signal Inversion: The Systematic Miscalibration of Deception Detection and Its Implications for Justice System Design

A. C. Applebee and A. D. Masters

---

Abstract

This paper presents converging evidence that the criminal justice system's methods for assessing credibility are not merely inaccurate but systematically inverted: the behavioural and linguistic cues that trained investigators, jurors, and the general public interpret as indicators of deception are empirically more strongly associated with truthful communication, while the cues interpreted as indicators of honesty characterise rehearsed or deceptive speech. Original analysis of the Belief-Reality Inversion Matrix, matching 23 behavioural cues between global belief data (Global Deception Research Team, 2006; N = 11,227; 75 countries) and meta-analytic empirical data (DePaulo et al., 2003; 158 cues; N > 10,000), demonstrates that 91.3% (21/23) of credibility cues are systematically inverted (binomial test, p 10,000 participants). Twenty-three cues were identifiable in both datasets with sufficient specificity to permit matching.

For each matched cue, the belief direction (whether the public believes the cue increases in liars) was compared to the empirical direction (whether the cue actually increases in liars, decreases in liars, or shows no reliable relationship). A cue was classified as "inverted" if the public believed it increased in liars but empirical evidence showed it either (a) had no reliable relationship to deception (d near zero) or (b) actually decreased in liars (i.e., increased in truth-tellers).

Results:

Of the 23 matched cues, 21 were classified as inverted. The inversion rate was 91.3%.

Nonverbal cues: 10/11 inverted (90.9%)

Paralinguistic cues: 5/5 inverted (100.0%)

Verbal cues: 6/7 inverted (85.7%)

Statistical test: Binomial test against the null hypot