Ask HN: Is it possible to create a image processing based lie detector?

2 points by sumitviii 2 years ago | 9 comments
Hello, Seeing how most people who guess well if someone is lying do it from 'reading a face', it can be said that there must be auditory and visual cues in a person's expressions that reveal if someone is lying. Is it possible to create a lie detector just based on reading and processing someone's facial expressions and voice?

This can have a lot of uses. Courts can know if a defendant is guilty or innocent immediately in almost all cases. Businesses can know if their suppliers and employees are capable. Employees can know if the employer will actually give them the raise that they have been promising. We may all be able to know if our dear politicians are lying.

This will improve our lives and it seems like a marketable product. Why hasn't anyone made it so far.

  • leephillips 2 years ago
    Do you have any evidence that there are people “who guess well if someone is lying”? In the absence of a controlled, reproduced study, I’ll stick with my default assumption, that this is just another rumor or superstition.
    • gus_massa 2 years ago
      Note that the effectiveness usual lie detector is quite dubious https://en.wikipedia.org/wiki/Polygraph#Effectiveness I expect this method to be also ineffective and used to make false accusations.
      • anigbrowl 2 years ago
        Sure, you can use a photocopier as a lie detector if your subject is sufficiently gullible. The real questions are why you think it's needed and what level of abuse/misapplication you're willing to be responsible for.

        If your use cases cases as described in the OP are for real rather than satirical, you are a naive person and should spend more time reading about the history of technology to understand how and why good intentions can have catastrophic outcomes.

        • nosmokewhereiam 2 years ago
          See also: Voice Stress Analysis.

          https://liarliar.sourceforge.net/

          • gvb 2 years ago
            https://www.technologyreview.com/2020/03/13/905323/ai-lie-de...

            Skip to the end for the tl;dr:

            The foundational premise of AI lie detection is that lies are there to be seen with the right tools. Psychologists still don’t know how valid that claim is [...]. The promise of a window into the inner lives of others is too tempting to pass up, even if nobody can be sure how clear that window is.

            “It’s the promise of mind-reading,” says Wilde. “You can see that it’s bogus, but that’s what they’re selling.”

            • joezydeco 2 years ago
              • smoldesu 2 years ago
                Nope! Would still be fun to try, though.
                • ninethirty 2 years ago
                  There's GitHub repos and research papers on this topic.
                  • booboofixer 2 years ago
                    You should pitch it to the governments running Xinjiang