Paradox of unanimity

- [Instructor] Hi, I'm Jill Butler, and this is Universal Principles of Design. In this movie, the paradox of unanimity, or when something is too amazing to be true. In April 2007, Michele Kiesewetter and Martin Arnold, two police officers in Heilbronn, Germany, were taking a lunch break in their patrol car. Without warning or provocation, two assailants approached their car from behind and shot them. Kiesewetter was killed, and Arnold was seriously injured.

This brazen attack shocked the nation and triggered one of the largest criminal investigations in German history. As police processed DNA evidence from the crime scene, they were stunned by the results. The DNA matched a notorious serial killer that had been dubbed the Phantom of Heilbronn, or the Woman Without a Face, a killer whose DNA had been recovered from over 40 crime scenes in Austria, France, and Germany over a period of 15 years.

From the DNA, police knew that the killer was female and of Eastern European descent, but that's about it. Police tried to create composite sketches from eyewitness accounts, but the Phantom's appearance varied widely across crime scenes. She seemed to be an assassin with a thousand faces, many of which appeared to be male. The papers declared the case the most mysterious serial crime of the past century. As the investigation escalated, a $400,000 bounty was placed on the Phantom's head.

Psychologists from around Europe were called in to create profiles. Police even consulted psychics and fortune tellers to help hunt her down. And this is what they found out. The Phantom defied traditional patterns and profiles of serial killers. She committed a wide range of crimes, from robberies to murders, with a seemingly random mix of accomplices. She seemed to radically alter her appearance for each crime, and she had committed her unprecedented crime spree over a wide geography and a period of time with impunity.

She seemed to be everywhere, and yet they basically knew nothing about her. And then, in March 2009, after almost two decades relentlessly pursuing this Phantom serial killer, the Heilbronn Police finally cracked the case. The DNA was traced to a group of women of Eastern European descent working in Austria at a factory that manufactured cotton swabs. The same cotton swabs, it turns out, used by forensic teams throughout Europe.

Yep, you guessed it. The DNA recovered from the crime scenes were already on the cotton swabs prior to being used for DNA collection. It turns out that although cotton swabs are sterilized in the production process, human cells from skin or sweat can survive. The Phantom of Heilbronn turned out to be a genuine phantom. She didn't exist. After this embarrassing revelation, German police closed the case on the Phantom of Heilbronn.

And so what, you may be asking, does any of this have to do with design? Well, the Phantom of Heilbronn illustrates a powerful but obscure phenomenon known as the paradox of unanimity. And it turns out that this paradox frequently leads us to misinterpret data and make bad decisions. The paradox of unanimity occurs when data are, in effect, too amazing to be true. That is, paradoxically, when large groups of people agree on something too much, or the data from complex systems are too consistent, or the results from analysis conform to expectations too well, then it is more likely to be evidence of systemic error, bias, or corruption than to actually be true.

Now the reasoning sounds counterintuitive, but researchers from the University of Adelaide and the University of Angers have demonstrated the paradox mathematically. For example, in police line-ups, the probability that the witnesses are correct begins to decrease after only three unanimous identifications. Counterintuitively, if one of the witnesses were to identify a different suspect, then the probability that the other witnesses were correct would substantially increase. When unanimity increases, confidence in the conclusion actually decreases.

So how can designers apply the paradox of unanimity in design contexts? Two ways, better data analysis and better group decision making. Remember the Volkswagen scandal where the company programmed their cars' computer chips to run in a special high efficiency mode to pass emissions tests? The fraud was discovered when experts realized that emissions from older Volkswagens were basically the same as their brand new cars. The consistency of the data was too amazing to be true, and that's how they got busted.

It's not uncommon for various kinds of user research, from interviews to usability testing to A/B testing, to return unanimous or perfectly conclusive results. The paradox of unanimity teaches us to be highly skeptical of such outcomes. They almost always indicate major error or bias in the system or methods. And finally, better group decision making. A trend in many organizations is to drive decisions to consensus, but the paradox of unanimity teaches us that true unanimity in most complex situations is highly unlikely.

And so driving groups to consensus actually undermines quality decision making. Good decision making practice accepts that there will be differences of opinion and supports mechanisms to record dissent and move forward with decisions. The alternative is artificial consensus, which helps no one. So whether you use your knowledge of the paradox of unanimity to identify systemic errors, bias, or malfeasance in systems, to improve the quality of your decision making and group governance, or to avoid spending 15 plus years hunting phantom killers, remember, when something is too amazing to be true, it probably is.