People Who Read People: A Behavior and Psychology Podcast Can behavior clusters help determine a lie? Tim Levine on deception detection science
Nov 13, 2025
Tim Levine, a leading deception researcher and author of 'Duped', delves into the science of lying and the myths surrounding deception detection. He critiques the common belief in nonverbal cues as reliable indicators of lies, discussing the limitations of combining cues and the inconsistencies in research findings. Levine emphasizes the need for careful evaluation of evidence and explores the role of context in understanding deception, while also expressing skepticism about the practical value of microexpressions and AI in this field.
AI Snips
Chapters
Transcript
Episode notes
Cues Don’t Generalize Well
- Nonverbal cues rarely generalize across people and situations, making single-cue detection unreliable.
- Tim Levine argues packages of behavior (demeanor) matter more than isolated signals for believability judgments.
Why Meta-Analyses Tell Different Stories
- Single-cue meta-analyses (DePaulo et al.) show most cues average to zero diagnosticity across studies.
- Hartwig & Bond's multi-cue results look stronger but can reflect publication bias and cherry-picked study-level effects.
Multiple-Cue Claims Can Be Misleading
- Hartwig & Bond aggregate study-level 'best' cues, which inflates apparent accuracy because different studies highlight different cues.
- Cross-study replication is necessary to show any cue package truly generalizes.