Lie-detecting software uses real court case data

December 10, 2015
Credit: George Hodan/public domain

By studying videos from high-stakes court cases, University of Michigan researchers are building unique lie-detecting software based on real-world data.

Their prototype considers both the speaker's words and gestures, and unlike a polygraph, it doesn't need to touch the subject in order to work. In experiments, it was up to 75 percent accurate in identifying who was being deceptive (as defined by trial outcomes), compared with humans' scores of just above 50 percent.

With the software, the researchers say they've identified several tells. Lying individuals moved their hands more. They tried to sound more certain. And, somewhat counterintuitively, they looked their questioners in the eye a bit more often than those presumed to be telling the truth, among other behaviors.

The system might one day be a helpful tool for security agents, juries and even mental health professionals, the researchers say.

To develop the software, the team used machine-learning techniques to train it on a set of 120 video clips from media coverage of actual trials. They got some of their clips from the website of The Innocence Project, a national organization that works to exonerate the wrongfully convicted.

The "real world" aspect of the work is one of the main ways it's different.

"In laboratory experiments, it's difficult to create a setting that motivates people to truly lie. The stakes are not high enough," said Rada Mihalcea, professor of computer science and engineering who leads the project with Mihai Burzo, assistant professor of mechanical engineering at UM-Flint. "We can offer a reward if people can lie well—pay them to convince another person that something false is true. But in the real world there is true motivation to deceive."

The videos include testimony from both defendants and witnesses. In half of the clips, the subject is deemed to be lying. To determine who was telling the truth, the researchers compared their testimony with trial verdicts.

To conduct the study, the team transcribed the audio, including vocal fill such as "um, ah, and uh." They then analyzed how often subjects used various words or categories of words. They also counted the gestures in the videos using a standard coding scheme for interpersonal interactions that scores nine different motions of the head, eyes, brow, mouth and hands.

The researchers fed the data into their system and let it sort the videos. When it used input from both the speaker's words and gestures, it was 75 percent accurate in identifying who was lying. That's much better than humans, who did just better than a coin-flip.

"People are poor lie detectors," Mihalcea said. "This isn't the kind of task we're naturally good at. There are clues that humans give naturally when they are being deceptive, but we're not paying close enough attention to pick them up. We're not counting how many times a person says 'I' or looks up. We're focusing on a higher level of communication."

In the clips of people lying, the researchers found common behaviors:

  • Scowling or grimacing of the whole face. This was in 30 percent of lying videos vs. 10 percent of truthful ones.
  • Looking directly at the questioner—in 70 percent of deceptive clips vs. 60 percent of truthful.
  • Gesturing with both hands—in 40 percent of lying clips, compared with 25 percent of the truthful.
  • Speaking with more vocal fill such as "um." This was more common during deception.
  • Distancing themselves from the action with words such as "he" or "she," rather than "I" or "we," and using phrases that reflected certainty.

This effort is one piece of a larger project.

"We are integrating physiological parameters such as heart rate, respiration rate and body temperature fluctuations, all gathered with non-invasive thermal imaging," Burzo said.

The researchers are also exploring the role of cultural influence.

"Deception detection is a very difficult problem," Burzo said. "We are getting at it from several different angles."

For this work, the researchers themselves classified the gestures, rather than having the computer do it. They're in the process of training the computer to do that.

The research team also includes research fellows Veronica Perez-Rosas and Mohamed Abouelenien. A paper on the findings titled "Deception Detection using Real-life Trial Data" was presented at the International Conference on Multimodal Interaction and is published in the 2015 conference proceedings. The work was funded by the National Science Foundation, John Templeton Foundation and Defense Advanced Research Projects Agency.

Explore further: The lying game

Related Stories

The lying game

October 30, 2015

University of Huddersfield investigative psychology lecturer Dr Chris Street is making breakthroughs that are leading towards a clearer understanding of how humans tell lies and how their deceptions can be detected. For more ...

New study finds group discussion improves lie detection

June 11, 2015

Though many people believe they can recognize when someone is lying, detecting deception is difficult. Accuracy rates in experiments have proven to be only slightly greater than chance, even among trained professionals.

Recommended for you

Volvo to supply Uber with self-driving cars (Update)

November 20, 2017

Swedish carmaker Volvo Cars said Monday it has signed an agreement to supply "tens of thousands" of self-driving cars to Uber, as the ride-sharing company battles a number of different controversies.

New method analyzes corn kernel characteristics

November 17, 2017

An ear of corn averages about 800 kernels. A traditional field method to estimate the number of kernels on the ear is to manually count the number of rows and multiply by the number of kernels in one length of the ear. With ...

5 comments

Adjust slider to filter visible comments by rank

Display comments: newest first

dogbert
3 / 5 (3) Dec 10, 2015
The continual desire to test people's veracity, despite the fact that we know such tests are not accurate, is disturbing.

We know, for example, that a psychopath can lie convincingly because the psychopath experiences none of the stresses normal people experience when lying.

We also know that honest people who are telling the truth can exhibit behaviors and physical responses which can be interpreted as lying when they are under stress such as when they are being subjected to a lie detector test.

Lie detector tests should be eliminated and any results obtained by such tests should be received with prejudice.
antialias_physorg
5 / 5 (3) Dec 10, 2015
"People are poor lie detectors," Mihalcea said. "This isn't the kind of task we're naturally good at.

If we were people wouldn't lie (or become more adept at lying to the point where most people are,again, bad lie detectors)

The continual desire to test people's veracity, despite the fact that we know such tests are not accurate, is disturbing.

Gains can be figuring out when your car salesman ist trying to BS you, figuring veracity of court testimony, or finding out if someone is trying to enter a country for legitimate reasons or to try and bomb it.

We know, for example, that a psychopath can lie convincingly because the psychopath experiences none of the stresses normal people experience when lying.

Not everyone is a psychopath. We need to distinguish error cases between false positives (truth being detected as a lie) and false negatives (lie being detected as a truth). If you can optimize one then your test is getting better (both would be best)
Captain Stumpy
3.7 / 5 (3) Dec 10, 2015
The continual desire to test people's veracity
@dog
because by continuing to test, we learn, and thus we can evolve better methods
kinda like this: https://en.wikipe...c_method

Lie detector tests should be eliminated and any results obtained by such tests should be received with prejudice
1- normally, lie detectors are not allowed in court cases, except as corroborating evidence (much like hearsay or other circumstantial evidence)
so, there is already much prejudice applied etc
https://en.wikipe...of_proof

https://en.wikipe...ce_(law)

https://en.wikipe...olygraph

2- polygraphs can be effective at focusing investigations or determining leads

just because a tool is not 100% useful doesn't mean it can't be effective
after all - most people have used a regular screwdriver as a chisel or other pry-tool in their life...
DavidW
1 / 5 (2) Dec 10, 2015
The first place to test this is on the people who created it for any question they can possibly have guilt for. And tomorrow too. And every day after. antialias should not use the truth, because antialias doesn't understand the word or value the truth itself. antialias is most important however, but chooses to use it to contest truth. We all know what that means.....a result of pain on other life and a broken mind. Most important is most important. Work it out dude. If you have 1/2 the intelligence you claim to have, a complete apology will be forthcoming to everyone that reads here, that you did indeed, push the most important truth in life aside as if it has no value, repeatedly and never once founded your position to do so in truth.. Now you see your words have no value until you accept. Unless of course, the value you speak of is to cause harm to life.
DavidW
1 / 5 (2) Dec 10, 2015
I have not been talking about something trivial ... I have stated, as clear as I think I can, "most important". Get a grip. Think for yourself. It should be possible to write some code to filter out unsubstantiated evidence and retesting the results that are substantiated on the first defined truthful "point" of reality: Life is most important in life. I would't leave that to the people in science who talk about this stuff at pleasure, while children are drowning and starving. Better for an equal to point out the error of another's logic, so that their equal can think sanely for themselves.

Please sign in to add a comment. Registration is free, and takes less than a minute. Read more

Click here to reset your password.
Sign in to get notified via email when new comments are made.