Beware of AI-based Deception Detection, Warns Scientific Community
Another problem they describe is the occurrence of “biases” in the decision-making process. The original hope was that machines would be able to overcome human biases such as stereotypes or prejudices. In reality, however, this assumption often fails due to an incorrect selection of variables that humans feed into the model, as well as the small size and lack of representativeness of the data used. Not to mention the fact that the data used to create such systems is often already biased.
The third problem is of a fundamental nature: “The use of artificial intelligence in lie detection is based on the assumption that it is possible to identify a valid cue or a combination of cues that are unique for deception,” explains Kristina Suchotzki. However, not even decades of research have been able to identify such unique cues. There is also no theory that can convincingly predict their existence.
High Susceptibility to Errors in Mass Screenings
However, Suchotzki and Gamer do not want to advise against working on AI-based deception detection. Ultimately, it is an empirical question as to whether this technology has the potential to deliver sufficiently valid results. However, in their opinion, several conditions must be met before it should be even considered to use in real life.
“We strongly recommend that decision-makers carefully check whether basic quality standards have been met in the development of algorithms,” they say. Prerequisites include controlled laboratory experiments, large and diverse data sets without systematic bias and the validation of algorithms and their accuracy on a large and independent data set.
The aim must be to avoid unnecessary false positives - i.e. cases in which the algorithm mistakenly believes it has detected a lie. There is a big difference between the use of AI as a mass screening tool, for example at airports, and the use of AI for specific incidents, such as the interrogation of a suspect in a criminal case. “Mass screening applications often involve very unstructured and uncontrolled assessments. This drastically increases the number of false positive results,” explains Matthias Gamer.
Warning to Politicians
Finally, the two researchers advise that AI-based deception detection should only be used in highly structured and controlled situations. Although there are no clear indicators of lies, it may be possible to minimize the number of alternative explanations in such situations. This increases the probability that differences in behavior or in the content of statements can be attributed to an attempt to deceive.
Kristina Suchotzki and Matthias Gamer supplement their recommendations with a warning to politicians: “History teaches us what happens if we do not adhere to strict research standards before methods for detecting deception are introduced in real life.” The example of the polygraph shows very clearly how difficult it is to get rid of such methods, even if evidence of low validity and the systematic discrimination against innocent suspects accumulates later.
