Detecting Deception with Disfluencies during Language Production in Hindi

You can send poster presenters your questions using this Google Form: https://forms.gle/LoxMiNTArxzHSi6w6

You can view the poster in more detail here: https://drive.google.com/file/d/1Rsbi...

Abstract: Spontaneous speech is rarely fluent. Disfluencies, including pauses, filled pauses (such as "uh" and "um"), repetitions, and false starts, are frequent. They reflect planning delays, working memory load, structural, lexical, or phonological complexity, and/or change in discourse.

For a speaker, lying carries cognitive effort. There is an inhibition of the truth and additional construction/execution of the lie, effectively affecting speech. Congruently, literature suggests that pitch and disfluencies serve as cues to deception.
Between two experiments in Hindi, we find that individuals use more filled pauses when requiring more processing/inhibition. In a naturalistic setting, the average filled pause rate was over 3 times higher in truth-tellers, while the silent pause rate was narrowly higher in liars. Conversely, in the false opinion task, filled pause rate was significantly different across conditions, indicating greater fluency when supporting a statement. There was no difference in pitch observed with respect to deception.

By Eashani Sharma and Shruti Sircar