Artificial Intelligence has no room for error!

Artificial intelligence is making its way into all areas of our lives at a phenomenal speed. Just consider the automobile. Tesla, the fourth largest automobile manufacturer in the U.S. thanks to its many innovations, has included an autopilot function in its vehicles. This function is aimed at mimicking a driver’s judgement and, in actual fact, is a smart autonomous driving feature and an intelligent speed regulation system. For example, the Tesla vehicle can brake on its own (without the driver touching the pedal) to avoid an object or can change lanes, moving either to the left or to the right, thanks to its integrated sensors.

Tesla vehicles have travelled roughly 200 million kilometres in autopilot mode to date. However, the company recently made the headlines for all the wrong reasons following its first fatal crash involving a driver who was using the autopilot function. One accident is all that is needed to shatter consumer confidence and for people to conclude that self-driving cars are dangerous and that we should not be relying on this technology. A person behind the wheel will always make the best decision. And yet, studies show that, on average, fatal accidents occur every 95 million kilometres when a person is behind the wheel while the very first fatality occurred with a self-driving car after 200 million kilometres!

Google Car is also conducting tests that would allow drivers to do other things, such as take a nap, watch a movie or read a book while their vehicle is driving on autopilot.

A recent University of Pennsylvania study concluded that people lose confidence in algorithms after seeing them make mistakes. Yet, research shows that algorithms are better forecasters than humans. Moreover, people also tend to reject outright algorithms providing results that largely exceed human forecasts, as though the algorithmic forecasts were simply impossible. In other words, algorithms are just not given a chance. They are damned either way, whether they err or perform exceptionally well!

Algorithms are now becoming an important part of different areas of our lives. Robotic financial advisors are threatening the jobs of financial planners and software that can analyze x-rays has been developed, replacing doctors who spent more than ten years learning this specialty. What about using algorithms to predict human behaviour?

Do you believe that human behaviour cannot be predicted because human beings are too complex? Make no mistake about this. Researchers at the University of South Carolina have developed a machine that is able to identify individuals showing symptoms of depression. Results show that a machine with voice recognition capabilities can accurately identify people showing signs of depression using algorithms that analyze a patient’s replies during a diagnostic interview with a physician. This represents a major breakthrough in psychology and psychiatry. It is a widely known fact that physicians often misdiagnose patients, believing that they are depressed when they are not. In fact, there are apparently three times as many false positives (i.e. people identified as being depressed when they are not) as there are false negatives (i.e. people not diagnosed as being depressed when they are). Doctors are apparently wrong half of the time. Extreme caution is used when making a clinical judgement regarding depression. Above all else, doctors want to avoid missing a patient who really is suffering from depression, since this could have major repercussions for the patient as well as doctors themselves.

With D-Teck, we have dared to consider using algorithms and smart machines in predicting human behaviour. Like our colleagues in other fields, who are toying increasingly with the idea of using artificial intelligence, we believe that this technology can be a powerful tool that can help us to assess people with increasing accuracy.

However, we will need to show more openness in terms of how we view algorithms to avoid throwing the baby out with the bath water when an algorithm makes a mistake. At the same time, algorithm developers must ensure that they take all the necessary precautions to manage the margin for error.

About Martin Cloutier, Co-Founder, Business Leader & Organizational Psychologist
As President of D-Teck, Mr. Cloutier sets the company vision and coordinates its execution. He is the cofounder of D-Teck, and his mission is to reinvent online psychometric tests. An organizational psychologist by training, his approach combines artificial intelligence with organizational psychology to help clients make better decisions about hiring, promotions and development.