Artificial intelligence is becoming more than a productivity tool. Increasingly, people are turning to AI systems for advice about money, careers, technology, relationships, health, and other personal problems.
New research highlighted by MIT Sloan School of Management suggests there is another important reason people may choose an AI adviser over a human one: AI can feel less judgmental.
Researchers found that people generally consider human advisers more competent. However, that preference can change when seeking advice requires admitting something embarrassing. The findings point toward an emerging role for AI that has relatively little to do with superior intelligence and much more to do with human psychology.
Key Points
- People generally preferred human advisers when AI and humans were presented as similarly competent, but embarrassment could shift that preference toward AI.
- AI may encourage people to disclose information they would otherwise hide because they perceive a computer system as less socially judgmental.
- The research reflects a broader real-world trend, but feeling comfortable disclosing information to AI does not mean AI is necessarily more accurate, confidential, or qualified to give professional advice.
The Competence Versus Judgment Trade-Off
The research, conducted by Eric So of MIT Sloan and Abigail Sussman and Fiona Yang of the University of Chicago Booth School of Business, examined what the researchers describe as a “competence-judgment tradeoff.”
People seeking professional advice must decide whether the expertise they expect to receive is worth the possibility of being judged for what they disclose.
Under ordinary circumstances, human advisers held an advantage.
Participants generally considered human professionals more capable than AI advisers in areas including financial, medical, technological, and career advice. When humans and AI were presented as similarly competent, participants also generally preferred humans and were willing to pay more for human advice.
Embarrassment changed the equation.
When participants had to disclose information that could make them feel ashamed, they became considerably more receptive to an AI adviser.
Embarrassment Changes How People Seek Help
One experiment involved 965 participants who were given hypothetical situations involving issues such as substantial credit card debt.
Researchers varied the cause.
In one scenario, the debt resulted from something largely outside the person’s control, such as unavoidable medical expenses. In another, the problem resulted from potentially embarrassing behavior, such as frivolous spending.
Participants then chose whether they would rather explain the situation to a human or an AI adviser.
Most preferred a human when embarrassment was low. When embarrassment increased, however, AI became more attractive.
A second experiment involving 744 participants manipulated both embarrassment and perceived adviser competence.
Making AI appear more competent increased people’s willingness to choose it. Increasing embarrassment produced a similar effect.
The implication is important: AI does not necessarily have to become universally superior to human professionals to attract clients. In situations where disclosure itself is uncomfortable, reducing the perceived social cost of asking for help can become an advantage.
People Explain Themselves Differently to AI
Perhaps one of the most interesting findings came from the third experiment.
Participants described real embarrassing financial or technological problems to either a human or AI adviser. Independent reviewers then examined how directly participants acknowledged what happened, the amount of information provided, and whether people attempted to justify their actions.
Participants offered justifications to human advisers 33% of the time, compared with only 15% when communicating with AI.
That difference suggests people may engage in a form of social self-protection when communicating with another person.
We do not simply report information to other people. We frequently explain why something happened, defend our decisions, minimize mistakes, or attempt to influence how another person perceives us.
A computer seemingly removes part of that social pressure.
AI Could Become the First Stop for Sensitive Problems
The findings could have significant implications for professional services.
Financial institutions, health organizations, technology companies, career services, insurers, and other businesses may increasingly use AI as an initial point of contact.
Someone embarrassed about excessive debt, a cybersecurity mistake, career difficulties, or another sensitive issue might disclose the problem to AI before speaking with a professional.
That could be beneficial if the alternative is avoiding help altogether.
AI can also operate continuously and at relatively low cost, potentially making basic information and guidance available to people who cannot easily access professional services.
But there is an important distinction between encouraging disclosure and providing reliable professional advice.
Is This a Realistic Picture of What Is Happening?
Broadly, yes, but the MIT findings should be interpreted as evidence of an important behavioral tendency rather than proof that people universally prefer AI for sensitive problems.
The underlying research consists of three controlled experiments. That makes it useful for identifying cause-and-effect relationships, but experimental choices and hypothetical scenarios do not automatically reproduce every factor involved in real financial, medical, legal, or career decisions.
Nevertheless, the central finding fits a larger body of research.
A 2024 literature review on self-disclosure to conversational AI found evidence that people can disclose more sensitive information when systems reduce social inhibition and concerns about evaluation. Research reviewed in that paper suggests that perceived judgment and the way an AI system is presented can significantly affect disclosure behavior.
There is also strong evidence that people are already turning to AI for highly personal subjects.
The American Psychological Association reported in 2026 that 77% of surveyed psychologists said their patients reported using AI, while more than one-third said patients were using AI as an additional source of mental-health support.
APA also notes that millions of people worldwide are using general-purpose AI chatbots and wellness applications for mental-health needs. Among the reasons these technologies appeal to users are accessibility, cost and the ability to reduce barriers associated with stigma and shame.
So the MIT research appears to be capturing a genuine phenomenon: AI can lower the psychological barrier to asking certain questions.
But there is another side to the story.
Feeling Safe Is Not the Same as Being Safe
A person may feel anonymous and unjudged when talking to an AI system, but that does not automatically mean the interaction is confidential.
AI services can have different policies governing how conversations are stored, reviewed, retained, or used. Users therefore need to understand a platform’s privacy policies before entering sensitive financial, medical, employment, or personally identifiable information.
There is also the question of accuracy.
The APA cautions that AI can deliver incorrect information confidently and should not replace qualified mental-health professionals. It recommends verifying medical or mental-health information with appropriate practitioners.
The World Health Organization has similarly emphasized that AI has considerable potential in health care while calling for governance, ethical safeguards, regulation, transparency, and appropriate human oversight.
These concerns extend beyond medicine.
A chatbot might make someone comfortable enough to admit a serious financial mistake, for example, while still misunderstanding tax regulations, lending requirements, legal consequences, or the person’s complete financial situation.
The psychological advantage identified by the MIT researchers therefore does not establish a competence advantage.
The Future May Be AI Plus Humans
The most practical model may not require choosing between AI and people.
AI could increasingly function as a confidential-feeling first interface where users organize their thoughts, describe uncomfortable situations, identify questions, and gather preliminary information.
Human professionals could then provide expertise, accountability, contextual understanding, and judgment when decisions become consequential.
In that model, AI’s greatest advantage may not be replacing the adviser.
It may be getting people through the door.
The MIT research points toward an important reason that could happen. People do not select advisers solely by asking who knows the most. They also consider whom they are comfortable telling the truth.
As AI becomes more capable and more deeply embedded in professional services, understanding that human factor may prove just as important as improving the technology itself.
Original Article
MIT Sloan — Clients prefer AI to human advisers when the details are embarrassing
Further Reading and Research
Eric So — AI Advisors and the Competence-Judgment Tradeoff in Information Disclosure
Springer — Self-disclosure to conversational AI: Literature review and research framework
American Psychological Association — Health advisory on generative AI chatbots and mental health
APA — Patients are bringing AI to therapy
APA — Guide to navigating AI-generated advice safely
World Health Organization — Artificial Intelligence for Health
SEO Elements
Yoast Key Phrase: AI advisers vs human advisers
Meta Description: Research finds people may prefer AI advisers for embarrassing problems because they feel less judged, but important risks remain.
Tags: artificial intelligence, AI advisers, AI chatbots, human advisers, generative AI, AI research, MIT Sloan, AI psychology, AI privacy, AI trust, human-AI interaction, AI professional services, conversational AI, AI trends
