Why stress is not proof of deception in an interview
Associate Professor Alex Kotov explains what behavioral analysis can observe and why interviewers still need context and judgment.
A pause or a change in tone may catch an interviewer’s attention. In a NoaMetrics conversation, Associate Professor Alex Kotov explains why those signals need context. They can suggest patterns for a person to examine; they do not prove that someone is lying.
Treat signals as probabilities
Kotov describes research into microexpressions, speech dynamics, and language patterns. The analysis looks for probabilistic indicators of behavior. It cannot turn one expression or pause into a reliable verdict about honesty.
Read audio, video, and language together
He names three approaches: acoustic analysis of speech, computer vision for visible behavior, and linguistic modeling of what a person says. Each contributes a different kind of observation; none can make the judgment for the interviewer.
Keep the interviewer responsible for interpretation
Stress does not equal deception. Kotov points to context and culture as reasons a signal may mean different things in different interviews. He describes AI as a tool for people to use, not a judge that should decide what a candidate meant.
Where the research might be used
Kotov mentions interviewer training, communication research, remote emotional-state analysis, and possible border or customs pilots. He presents these as areas of use or exploration. His team is working toward lower bias and better cultural modeling.
Four points from Kotov’s interview
- 01
Treat changes in expression, speech, or wording as clues, not conclusions.
- 02
Check what audio, video, and language analysis each observed.
- 03
Consider stress, context, and culture before interpreting a signal.
- 04
Leave the judgment with a person who can review the interview.