Sociale

The study showed that we are trusting computers more than humans

The study showed that we are trusting computers more than humans

Do you often ask for the calculator, even for the simplest mathematical calculations? These days there are many concerns that technology, such as artificial intelligence, is too smart for its own good.

Despite fears of how intrusive its algorithms are becoming, a new study finds that people are actually more willing to trust a computer more than their friend. Researchers at the University of Georgia in the US say this is especially true when people see tasks as too challenging to perform on their own.

However, these are not the only "difficult tasks" for which people turn to computers for help. From choosing the next song to listen to, to finding the most suitable pants, algorithms are gaining more and more weight in the daily decisions in people’s lives, regardless of whether they understand it or not.

"Algorithms are able to do a large number of tasks, and the number of tasks they are able to do is expanding virtually every day," said Erik Bogert, Ph.D. of Management, in a university publication.

"It seems like there is a bias towards greater reliance on algorithms, when a task becomes more difficult, and that effect is stronger than the bias towards relying on other people's advice," he adds.

The researchers evaluated the responses of 1,500 individuals in charge of counting people in a series of photographs. The team also gave participants suggestions on how to do this, from other people or with computer algorithms.

As the crowd in the picture grew larger and harder to count, volunteers were more likely to turn to computer suggestions than to follow their intuition. The study's co-author, Aron Shekter, says counting tasks are a perfect way to measure trust in computers.

As the number of people increases after each photo that passes, the task becomes more difficult for people to perform. The team of researchers adds that these tasks are also the type of problems that people expect to be solved by computers.

"This is a task that people perceive a computer would be better able to solve, even though it may be more subject to prejudice than counting objects," explains Shekter. But just because a software program is simply a collection of data does not mean that prejudice does not exist.

Shekter notes that face recognition and employment algorithms have recently been criticized because of the cultural biases involved in their programs. And they can lead to inaccuracies when matching faces with identities, or when verifying qualified candidates for a job.

This study is part of a larger Shekter study on US-funded human-machine cooperation. "The ultimate goal is to see groups of people and machines making different decisions, and find ways we can make them trust each other, and how that changes their behavior," Shekter concludes.

* Received from Bota.al