AI for lie detection Flashcards
Give examples of the practical applications of lie detection
Criminal investigations, insurance fraud, terrorism etc
What kind of lie detectors are in use now? (4)
Voice stress analysis (pitch, tremors), language analysis, physiological measures such as pitch and heart rate are used to construct polygraph tests which are used in a number of countries such as the US, UK and Belgium, screening passengers by observation at the airport after 911.
What do each of these have in common?
Cool acronyms and name and these methods often claim extremely high levels of accuracy however these claims have rarely been reflected in empirical research.
Give an example of a experimental design meant to measure lying in the lab
Participants are split into a lying and non-lying group. Liars leave the lab after receiving instructions and ‘commit a crime’ e.g steal an exam. Controls leave and do an innocent assignment. They then come back to second blind experimenter who interviews them and tries to figure out whether they are lying or telling the truth, both are instructed to give the same story.
Give some critiques of this kind of paradigm
- There is nothing at stake when ‘lying’
- lower motivation to attempt to hide a lie
What is a benefit of this kind of lab study?
They have control and know with absolute certainty who is lying and who is not.
Describe the alternative to lab research
Field research, for example going somewhere where polygraphs are being done, requesting the charts with physiological recordings from the police and then looking for a criterion which determines whether the polygraph was correct or not (e.g whether the person is convicted or not.)
What is the cost of this kind of research
You never know for certain who is lying and who is telling the truth
What problems arise with the existence of both these lines of research?
Often only lab or only field research is considers in meta analyses and literature reviews and the other on is ignored, this can lead to two very different conclusions.
Aside from claiming high accuracy and fancy acronyms, what underlying mechanism do these lie detection methods have?
Stress and distress
What is the problem with this underlying mechanism?
Many different variables can heighten our arousal, emotional valence and stress and cause a false positive. Questions about a sensitive case for example would cause more distress for someone accused than control questions.
People may also lie without feeling stress or arousal.
What is an alternative approach to detecting lying?
Lying is more difficult than telling the truth, it ‘demands invention, dissimulation, and a good memory.’ This is backed up with research through methods such as self report and reaction time (takes longer to lie).
What have brain imaging studies regarding lying revealed?
Study involved one group stealing an exam, one stealing 50 euro and then being asked about each crime. Activity in the right inferior frontal cortex, which has been associated with inhibition, has been associated with lying. This could suggest that the first intuition is to tell the truth, but they have to inhibit that and tell a lie instead.
Is lying always more difficult than telling the truth?
No, lying can be developed almost as a skill so that it becomes just as easy or easier than telling the truth.
If you want to use a cognition based approach to lie detection, what steps are advised to go about it?
- Use cognitive clues; e.g reaction time
- Elicit cognitive cues (as they are often hard to detect otherwise)
3.