Humans Can Trust Algorithms, But Completely?
Can humans trust algorithms’ judgments for all types of tasks?
Today, more and more companies are jumping on the bandwagon of big data. They are recruiting skilled data professionals to create algorithms and optimize recommendations. AI-driven leadership and educational programs are helping to identify and nurture top talent. However, many people are wary of algorithms and find it difficult to trust algorithms over human expertise or gut instincts.
According to one research in 1950 by Harvard Business Review, algorithms are outperforming humans. We have evidence already wherein machines and algorithms are used to detect cancer and deliver better patient care, overlook workers safety in manufacturing units, predict weather conditions, etc.
However, a lot of controversies surrounded this study because our egos forestalled us to accept that a machine can perform better than human expertise. To avoid such controversies yet adopting AI-driven solutions, companies are putting human faces on bots, introducing personal assistants, etc., highlighting human elements wherever possible.
Over time, things have changed. According to a new study by researchers from the University of Georgia, humans trust algorithms more than their peers, particularly in tedious tasks. In this study, 1500 respondents were shown photos and were asked to analyze the number of people in them.
The images that were shown initially involved 15 people, which gradually went on to grow to 5000 people to increase the complexity of the assignment. Hence, the respondents eventually relied on an algorithm to count the number of people. Machines are better at tedious tasks like counting than humans.
Nonetheless, the researchers accentuated that human viewpoints towards the accuracy of an algorithm is a significant factor. Giving the task to a machine unintentionally creates the chance of bias to crawl in without being known to the human participants.
The context of decision-making also plays an important role in how people perceive algorithms. One study revealed that when humans see algorithms making a mistake, they aren’t likely to trust them. Now this, hurts their accuracy. Besides, another study found that humans like listening to jokes from a close person over an algorithm, even though the algorithm is better than a human. But on the other hand, humans are likely to trust algorithms on moral decisions when self-driving cars or medicines are involved.
While the new study is only applicable in terms of tedious assignments and not other tasks. The study states that algorithms can be useful in a credit scoring or granting loans. However, algorithms don’t consider social factors when evaluating credit scoring, which is an important factor that humans are best at.
Although humans are agreeable with the pros of algorithms over human expertise, their faith in algorithms tends to decrease when they compare algorithms to their own judgement. So, human trust algorithms over someone else’s judgment but will choose their own judgement when it comes to choosing between their own and algorithm’s judgment.
While this new study clearly highlights that we can trust algorithms in certain tasks, it is still not sure if humans can practice that in reality.