“As the amount of data goes up, the importance of human judgment should go down,” argued Andrew McAfee in his Harvard Business Review blog post about Convincing People NOT to Trust Their Judgment, which is what he sees as the biggest challenge facing big data.
“Human intuition is real,” McAfee explained, “but it’s also really faulty.” He cited a variety of research documenting the poor performance of human experts against computer algorithms, noting “when experts apply their judgment to the output of a data-driven algorithm or mathematical model (in other words, when they second-guess it), they generally do worse than the algorithm alone would.”
The performance of human experts against computer algorithms was also examined by Daniel Kahneman in his book Thinking, Fast and Slow. Across a broad range of studies, which examined the accuracy of our ability to predict things such as academic performance, criminal recidivism, longevity of cancer patients, diagnosis of cardiac disease, credit risk, career satisfaction, suitability of foster parents, winners of football games and the future prices of Bordeaux wines, the accuracy of experts was matched or exceeded by a simple algorithm.
Complexity and consistency
Kahneman explained that one of the reasons experts are inferior to algorithms is that “experts try to be clever, think outside the box, and consider complex combinations of features in their predictions. Complexity may work in the odd case, but more often than not it reduces validity.”
“Simple combinations of features are better,” Kahneman argued. “Experts feel they can overrule the algorithm because they have additional information about the case, but they are wrong more often than not.”
Those findings gave me pause for thought since one of the lauded benefits of big data analytics is being able to leverage more information in our decision making.
Kahneman discovered that another reason for the inferiority of expert judgment is that humans are incorrigibly inconsistent in making summary judgments of complex information. When asked to evaluate the same information twice, they frequently give different answers. A review of separate studies on the reliability of judgments made by auditors, pathologists, psychologists, organizational managers and other professionals revealed that they contradicted themselves 20 percent of the time when asked to evaluate the same case on separate occasions.
By contrast, algorithms “do not suffer from such problems,” Kahneman explained. “Given the same input, they always return the same answer.”
Truth and consequences
“The practical conclusion,” McAfee argued, “is that we should turn many of our decisions, predictions, diagnoses, and judgments—both the trivial and the consequential—over to the algorithms. There’s just no controversy anymore about whether doing so will give us better results.” Of course, many people would disagree, which is why McAfee wrote a follow-up post about When Human Judgment Works Well, and When it Doesn’t.
“The aversion to algorithms making decisions that affect humans,” Kahneman explained, “is rooted in the strong preference that many people have for the natural over the synthetic. Asked whether they would rather eat an organic or a commercially grown apple, most people prefer the natural one. Even after being informed that the two apples taste the same, have identical nutritional value, and are equally healthful, a majority still prefer the organic fruit.” Our prejudice against the synthetic is significantly magnified when the decision is more consequential than picking an apple. Many proponents of algorithms, however, have strongly argued that it’s unethical to rely on intuitive judgments for important decisions if an algorithm is available that will make fewer mistakes.
“Their rational argument is compelling,” Kahneman explained, “but it runs against a stubborn psychological reality: for most people, the cause of a mistake matters. The story of a child dying because an algorithm made a mistake is more poignant than story of the same tragedy occurring as a result of human error, and the difference in emotional intensity is readily translated into a moral preference.”
That’s the ethical quandary facing algorithm-driven, evidence-based medicine. Though an algorithm makes fewer mistakes, one death by an algorithm’s mistake is somehow seen as worse than 100 deaths by a human’s error.
What say you?
Are you smarter than an algorithm? How comfortable are you with our increasing reliance on algorithms?