Clients prefer AI to human advisers when the details are embarrassing

When and why do clients prefer to use artificial intelligence rather than a human adviser? It’s a question attracting much attention as AI advisers flood the marketplace. A recent study from the MIT Sloan School of Management introduces a new factor into the decision-making process: embarrassment.

The new research shows that professional services clients who feel ashamed are more likely to prefer an AI assistant, even when they view a human alternative as more competent. The source of embarrassment can be anything from of racking up frivolous credit card debt to dealing with malware accidentally downloaded from an adult website.

“When we seek advice from other people, we’re often reluctant to share embarrassing details because we have a fear of being judged,” said Eric So, an MIT Sloan professor of global economics and behavioral science. “What our research shows is that people are more likely to turn to AI models to get professional advice when they’re embarrassed, because the model doesn’t judge them.”

The paper, “AI Advisers and the Competence-Judgment Tradeoff in Information Disclosure,” examines when people are more likely to open up to a human adviser versus an AI adviser. It was authored by So, along with Abigail Sussman and Fiona Yang from the University of Chicago Booth School of Business.

In the study, people consistently rated human advisers as more competent than AI advisers when evaluating who could best solve their problem across domains that included financial, technological, medical, and career advice.

But that situation changed once shame entered the picture.

• Across three experiments, when both human and AI advisers were described as having comparable competence, most people preferred human advisers and were willing to pay more for their services.
• That preference reversed when a client’s problem would require them to reveal embarrassing information.
• Study subjects also preferred AI when it was portrayed as being more competent than a human adviser.

How the researchers conducted their study

In one experiment, the researchers presented 965 participants with a hypothetical scenario, such as having a lot of credit card debt, and told them it was caused by either:

• Something beyond their control, such as unavoidable medical bills.
• Something embarrassing, such as frivolous spending.

Researchers asked the participants whether they would want to explain the problem to a human or to AI and then asked them to rate the extent of their embarrassment. The majority of people preferred to work with humans, except when the scenario was embarrassing.

In another experiment, the researchers tested the effects of two factors at once with a group of 744 participants:

• The scenario’s level of embarrassment, which researchers manipulated (low vs. high)
• The competence level of AI and human advisers, as rated and described by independent experts 

Presenting AI as highly skilled pulled people toward that option, as did a high embarrassment factor. If the researchers made the topic embarrassing enough, such as being caught overstating one’s professional qualifications, or made AI seem competent enough, people would choose AI over a human.

Artificial Intelligence for Financial Services

In person at MIT Sloan

Register Now

People feel the need to justify themselves to humans but not to AI

In the third experiment, people were asked to describe their own embarrassing situation regarding a financial or tech problem to a human or AI adviser. They were incentivized to be honest and were paid more for disclosing more about their situation. Two independent reviewers read the responses and scored:

• How directly people admitted the cause of their embarrassment.
• How much detail people gave.
• Whether participants felt the need to justify their behavior.

People offered justifications more often to humans (33% of the time) than to AI (15% of the time).

People weigh multiple factors when choosing an AI or human adviser

Overall, the experiments show that the choice between using a human or an AI adviser comes down to a trade-off. People weigh an adviser’s competence against the risk of being socially judged. 

This combined “competence-judgment gap” — that is, evaluating whether the benefit of expertise outweighs the cost of social judgment — is a better indicator of whether someone will disclose their problem than either factor alone, the study concluded.

There’s room for AI chatbots to gain market share

The study highlights the value of using AI chatbots in certain advisory positions, So said.

“We predict that AI advisers will really take off in contexts where embarrassing situations are more prevalent and where the models are perceived as being equally or more capable than humans,” So said. 

Specifically, the presence of an AI option can spur people in embarrassing situations to seek help — and potentially achieve a better outcome. 

Diagnosing the root cause of a problem is often key to preventing it from recurring, So noted. “But if you aren’t willing to share it in a candid way with people, then you may not be able to get advice that actually solves your problem,” he said. “AI offers significant promise in getting people advice for less money on a platform where they feel less judged.”

The choice of adviser needn’t be an either-or situation. A successful strategy can involve both AI and human interaction, So said.

Transparency builds trust — for human advisers and chatbots alike

Companies should give customers a choice about how to interact with their advisory services, So suggested, by allowing them to choose to share sensitive information with a human or with an AI chat system — whichever way they feel less judged.

That said, So stressed that companies must be transparent around how customers’ data is used. If someone discloses something embarrassing to an AI system but is still worried that the information will ultimately be seen by a human, that defeats the appeal of AI’s nonjudgmental advice.

The study’s results also have a takeaway for human advisers: They should remember to reiterate to clients that they, too, conduct business in a nonjudgmental, safe space.

“As AI models become more capable, elements of human touch are going to be quite important,” So said. “One of the major critiques of AI systems is that they can understand the narrow confines of a problem but they miss the broader context. Humans can offer trust and continued support.”

Eric So is the Sloan Distinguished Professor of Global Economics and Behavioral Science at the MIT Sloan School of Management, faculty co-director of the AI Executive Academy, faculty chair of MIT Sloan’s PhD program, and lead faculty for the MIT Sloan Generative AI for Teaching & Learning hub. His current research portfolio spans interconnected topics, including artificial intelligence, behavioral economics, human-computer interactions, and regulatory policy. His book “The Collision: What AI Does to Us” will be published in October 2026.

Abigail Sussman is a professor of marketing at the University of Chicago Booth School of Business. She researches how individuals form judgments and make decisions, investigating questions at the intersection of psychology, economics, and finance. Her central research examines psychological biases that can lead consumers to commit errors in budgeting, spending, borrowing, and investing, with the aim of improving financial well-being.

Fiona Yujin Yang is a research professional at the University of Chicago Booth School of Business.