People Don’t Just Distrust ChatGPT. They Judge the Person Who Uses It
Eight studies found that people see those who use technology to get advice as less warm and caring. This could affect hiring, feedback, and how people use technology at work.
EXPERT OPINION BY NICK HOBSON, Managing Director of North America, Influence at Work, and Behavioral Science Adviser @NICKMHobson
Photo: Adobe Stock
Imagine this situation: A university student needs a list of books and other reading materials for an essay about early art. In one version of the story, he asks a professor for advice. In another, he asks ChatGPT. He then wants to collect public feedback through short surveys, with no payment for taking part. Strangers are asked how many surveys they would be willing to complete.
People were more likely to agree to take part in the survey when they were told the reading list came from a professor. The student, the essay, and the request were exactly the same. But he lost about as much goodwill as one survey's worth simply because he had got the reading list from a different source.
This experiment was one of eight studies in a research paper by Jianning Dang and Li Liu, published in the Journal of Personality and Social Psychology. Psychologists have spent years studying “AI aversion.” This means people tend to distrust an algorithm's judgment even when it is as good as, or better than, the judgment of a human expert. Dang and Liu asked the next question: Does this distrust also affect how people judge someone who uses the tool? Their studies, which included 2,317 participants in the U.K. and China, found that it does.
What they measured
The researchers used much the same method in each study. Participants read a short story about someone who asked either an AI tool or a human expert for advice. The people in the stories included a hiring manager reviewing video interviews, someone choosing a wedding gift for a friend, a student, a patient, and a primary school teacher preparing reading materials. Participants then rated the person using two groups of traits linked to how human someone seems. “Human nature” traits, such as being friendly, sociable, and trusting, are qualities that set people apart from machines. “Human uniqueness” traits, such as being rational and mature, are qualities that set people apart from animals.
Across all the studies, people judged someone more harshly when that person asked AI for advice instead of asking a human expert. For example, in the wedding gift study, the person who asked ChatGPT for gift ideas received a lower score for how human they seemed. In most studies, the difference was greater for emotional qualities than for rational ones.
The first study showed this especially clearly. In it, 200 participants in the U.K. wrote their own descriptions of a hiring manager named James. They described what he thought, felt, and intended to do. People who read that James had used a computer program to help with hiring used words such as “feel” and “care” less often. But they used words such as “plan” and “think” just as often as the other group did. This suggests that people believed someone who used AI could still think and reason, but they were less certain about that person's feelings and emotions.
Participants also liked the person who used AI less. However, the difference in how human they thought the person was remained even after the researchers took those feelings of liking into account. The difference also remained no matter how much participants knew about AI or how much experience they had using it. This means that a lack of familiarity with AI is unlikely to explain the results.