Together with Ronald Poppe, Paul Taylor, and Gordon Wright, I took a plunge and tested our automated lie detection methods in the real world.
We invited 3 people to take part in this real world experiment that consisted of 2 parts. Mo, the 37-year old advertising consultant who lied to her parents for years about her sexuality. Ruth, the 46-year old parish priest, who feels it’s more important to be friendly than to be honest. And Ehiz, the 19-year old student and YouTuber, who lies frequently and quite instrumental. He also created an online alter ego who is living an extravagant lifestyle full of fancy clothes and parties.
First, we wanted to test how well our lie detection techniques worked outside the lab. For this purpose, we brought in researchers with three different specialties: nonverbal behavior, automatically measured using motion capture suits; language use, recorded, transcribed, and automatically analysed using the software program LIWC; and physiological responses, automatically measured using a smartwatch. We combined these three lie detection methods into one multimodal system and used it to distinguish between the truthful and deceptive behaviors of our participants. We started by collecting an extensive baseline for each participant. They told several truths and lies, and we captured their associated behaviors in a personal model for each individual. Next, they wore the kit for a full day, living their normal lives. Going to work, to the pub. Talking to family, friends, and strangers. During that day, they had several interactions with people they knew and strangers. We analysed the truthfulness of these interactions, selected the most truthful and dishonest ones based on the data from the three lie detection methods combined, and reported our conclusions back to the participants. All of the deceit examples we discussed, were confirmed by the participants, showing that our multimodal deception detection method also seemed to provide results when applied in the real world.
In the second part of the experiment, the three participants were fitted with the smartwatch, referred to as the ‘truthwatch’ by our participants, and were sent off to live a week without lying. During this week, they were asked to record their experiences in various ways. They could timestamp using their smartwatches whenever they were having a difficult moment, for example because of social backlash to their honesty, temptations to lie, or even when ending up lying. They could make video diaries to elaborate on their experiences, and were asked to fill out a written diary every night. During one of these days, they were filmed as well. I found this part of the experiment very fascinating, because it showed how deeply ingrained lying is in our society and in ourselves. Regardless of how hard they tried, all three participants lied several times during this week, and faced different difficulties. On the one hand, Mo experienced the benefits of being honest and got on a bit of high trying out this new version of herself. On the other hand, it also made her realize she was more dishonest than she previously thought, leading to her doubting the quality of her relationships and who she was as a person. Ruth often avoided to tell explicit lies by answering a different question. During this week, she was encouraged to stop using that tactic and be completely honest. As a result, she explored the thin line between being honest and saying everything that pops up in your mind. This week made her realise that a bit of deception is preferable to hurting people’s feelings by being to blunt, not just with strangers but also in a romantic relationship. Ehiz’ experience was very different from the two other participants. He had difficulties completing the challenge and called in sick during his week of living truthfully. He did not respond for a couple of days and decided not to upload an honest youtube video. He seemed to find it difficult to tell the truth, even when discussing his attempt to live truthfully with the experts.
This documentary has provided us with unique insights into our lying behavior. How often do we lie, when, to whom, and for what reasons? A major conclusion for me was that people lie much more often than we report, and maybe even than we realise we do. That means that we’re in need of new ways and methods to measure deceptive behavior. We also learned what difficulties arise when applying lie detection methods designed and tested in a lab, to the real world. Normally, researchers conduct lie detection experiments under very controlled circumstances. The liar sits at a table, with no distractions and nothing else to do then talk, and maybe lie, to the interviewer. Such a setting is great for determining lying behavior, since most behaviors that are shown, are a direct consequence of telling a lie. But the real world is messy and we often do several things at the same time. You have dinner with your family, or drinks in the pub with your friends. During these conversations, you may pick up our fork or glass whilst telling the occasional white lie, and all these behaviors add noise to the deception signal. With this experiment, we showed for the first time, that we can detect deception in noisy, real world, situations. We were very excited about these results, because even though our participant numbers were much too low to draw any scientific conclusions, it is a first indication that we’re doing something viable, and it has given me many ideas for future research that can help to further bridge the gap between research and practice.
The documentary was broadcasted on Wednesday the 29th of August on BBC 2, as part of the Horizon series. Preceding the broadcasting, it was tipped by The Sunday Times as the Critics’ Choice to watch.