腦科學測謊

Disception is an interesting topic because a successful lie most likely involves a combined effort of many cognitive faculties, like a well-orchestrated piece, in order to pull it off. And when you think about it, there are many differernt types of lies too; there are differences in motivation (selfish lie or altruistic lie, white lie), techniques (denial, fabrication, half truth, "I don't knows"), and sometimes, you can even lie with the truth (with good theory of mind, or with exaggeration to give it a false appearance)! 

This is perhaps why detecting lies, even with modern neuroscience techniques, is not easy. However, from a cognitive psychology standpoint, detecting the cognitive processes that are critical to successful lying is not too hard, and therefore is likely a more fruitful approach. Indeed, studies from the early 1990's have utilized what is now called the "guilty knowledge test" (GKT, or CIT for concealed information test) to take advantage of the fact that denial of existing memory traces (e.g., a seen knife) is slower and observable via EEG recordings. We conducted a study to investigate whether rotated object that is different from the culprit's memory would make a difference in GKT accuracy. To our surprise, the answer is no, at least not for familiar everyday-object (we used shoes of different kinds). Therefore, it seems that the forensic team does not have to worry too much about mimicing the original seen angle when choosing a photo for GKT.

Another research project we are pursuing is asking what can we do after a person is caught (and confessed) by GKT. For example, during a police interrogation, it is quite common for the suspects to make up false stories or misidentify seen objects in order to mislead further investigation. One possible way to counter such deceptive behavior is to measure and track working memory (WM) load, since fabricating coherent lies likely involve heavy WM usage. In one study (poster below), we designed a memory-based lying task to probe the roles of WM in the preparation and execution phase of lying. Participants memorized a list of words in the study session (i.e., old words), and had to use these old words to provide misleading answers when cued later in the testing session. Our behavioral results showed that people needed more time to make a deceptive response during the execution phase, and such prolonged deceptive RT was negatively correlated with each participant’s WM capacity. Our ERP findings were twofold. First, by contrasting lying and truth conditions, we observed a more negative-going frontal amplitude during the preparation phase (500~900 ms after cue onset), suggesting that WM preparatory processes can be detected long before a deceptive response is verbalized. Second, we observed a larger positive frontal and central amplitude during the execution phase (300~550 ms after target onset), and such amplitude was negatively correlated with participants’ lie-and-truth reaction time differences, suggesting that participants’ efficiency in producing deceptive responses can be readily traced electrophysiologically. Together, these findings suggest that WM capacity and preparation are crucial to efficient lying, and investigators can take advantage of such phenomenon, as well as its related eletrophysiological signatures, to uncover and detect deceptive behaviors.

For more details, please take a look at the following publications:

Tseng, P., & Cheng, T. (2025). Artificial intelligence in lie detection: Why do cognitive theories matter?. New Ideas in Psychology, 76, 101128. [WEB]

Hsu, A., Lo, Y. H., Ke, S. C., Lin, L., & Tseng, P. (2020). Variation of picture angles and its effect on the Concealed Information Test. Cognitive Research: Principles and Implications, 5(1), 1-18. [PDF]

Lo, Y.H., & Tseng, P. (2018). Electrophysiological markers of working memory usage as an index for truth-based lies. Cognitive, Affective, and Behavioral Neuroscience, 18, 1089-1104. [PDF]