Researchers from the University of Georgia found that people are more likely to trust artificial intelligence (AI) than humans, especially if a given task is too challenging for the human to complete. This study was part of a larger body of work that focused on analyzing how and when people work with algorithms to process information and make decisions.
People rely on the advice of algorithms every day to make decisions and streamline their lives, even if they don’t realize it. This study found that there are situations in which human reliance on algorithms is actually detrimental.
Source: iStock
The study had 1,500 participants who were asked to evaluate the number of people in photographs of a crowd. Participants could rely on suggestions generated by a group of other people or an algorithm. As the number of people in a crowd expanded and the counting got harder, participants were more likely to follow suggestions from the algorithm than count themselves. People expect computers to be good at this kind of task.
But reliance on AI comes with its own problems. AI is problematic when used for subjective tasks, like awarding credit or approving someone for loans. These tasks involve a lot of numbers but also require objective opinions based on social factors. Dependence on AI for these kinds of tasks leads to discriminatory practices because social factors are not taken into consideration.
Facial recognition and hiring algorithms have been scrutinized because their use revealed cultural biases embedded into the way that algorithms are built. This causes inaccuracies when matching faces to identities or screening qualified job candidates. Biases may not be present in simple tasks, but their presence during other tasks means it is important to understand how people rely on algorithms. Humans shouldn't rely on current AI for difficult tasks because the biases in these algorithms could actually have negative effects on the task results.
The team’s eventual goal is to look at groups of humans and machines making decisions, find how to get them to trust each other and how this changes behavior, starting with the fundamentals.
This study was published in Scientific Reports.
