What The Crowd Tell Us About When We Might Lie

Lying isn’t a good thing, but despite clear moral issues I suspect many of us tell fibs now and then.  A recent study led by the Mathematical Sciences Research Institute in Berkeley explored the kind of circumstances that prompt us to lie.

The crowdsourced research set out to tempt participants to lie in various ways in order to benefit them to varying degrees.  The test was setup on the Mechanical Turk platform, with users shown a randomly generated series of numbers between 1 and 90, whilst also being shown a single number that signifies a position on the list, with the user then asked to guess which number appears at this position.

What the participants don’t know is that all are shown the same series of numbers (25, 3, 63, 54, 28, 70, 37, 36, 26, 31, 43, 15, 30, 60, 33, 37, 15, 63, 16, 50, 4, 71, 79, 2, 85, 48), with half of the group asked to identify the number in the 19th position, and half the 22nd position (16 and 71 respectively).

So far so standard. Where things get interesting however is the reward players receive for a correct answer.  Those who answer 16 will only get 16 cents, but if they select the number either side (63 and 50), they could earn more.  It’s a fairly simple, if dishonest, strategy to take.  The incentive to cheat is reduced for those asked to select the 71, as their payout is closer to the maximum.

Towing the line

In total, around 800 people completed the experiment, and the results present an interesting insight into how and why we tow the line versus bending the rules.  Whilst 84% of people were found to be honest, a much higher proportion of those in the ‘unlucky’ group lied than their peers in the ‘lucky’ group.

The results suggest that we tend to calculate the likely payoff from our lie, and tend to do so if telling the truth will lead to us being worse off.  Of course, we shouldn’t besmirch the 50% of so of people that were found to tell the truth regardless of the impact of doing so.  Their unconditional honesty is to be applauded.

The participants tended to fall into one of three groups:

  1. Those who tell the truth regardless of their expected payoff (approximately 50% are thus)
  2. Conditional liars who do so when the payoff is sufficient.  Around 35% of people fall into this group.
  3. The final group are those will lie regardless, and they make up the remaining 15%.

Perhaps not that surprisingly, the women in the study were consistently more honest than the men, whilst dishonesty tends to take more time than honesty.

Automating lie detection

Help may be at hand from automated systems however.  Last year I wrote about technology being developed by San Diego State University to autonomously check for lies.  The device, called the Automated Virtual Agent for Truth Assessments in Real Time (AVATAR) is designed to provide accurate lie detection services to border staff.

“AVATAR is a kiosk, much like an airport check-in or grocery store self-checkout kiosk,” the team say. “However, this kiosk has a face on the screen that asks questions of travelers and can detect changes in physiology and behavior during the interview. The system can detect changes in the eyes, voice, gestures and posture to determine potential risk. It can even tell when you’re curling your toes.”

Passengers engage with the kiosk, which will ask them a number of questions, all the time monitoring the passenger in various ways in the hunt for tell-tale signs of lying or discomfort.  If they believe an individual warrants further inspection, they raise a red flag and human agents quiz them further.

The system is capable of detecting a wide range of data points from each interviewee, and therefore hopefully provide accurate predictions as to their honesty.  The overall plan is for the system to be used in a range of scenarios where honesty matters.