When Risk Feels Like Help, That’s the Risk

When Risk Feels Like Help, That’s the Risk



When Risk Feels Like Help, That’s the Risk

Last night I was in a room of therapists, researchers, and digital mental health people in the Bay Area. Smart, interesting people. One thing kept surfacing across every conversation: We are playing catch-up.

Research on AI and Human Psychology

The research on how artificial intelligence (AI) actually shapes human psychology is still forming, but we have early signals and cause for professional concern. A Peruvian study found that dependence on AI correlated with anxiety and depression in medical students (Sosa & Huancahuire-Vega, 2026). A Stanford study found that exposure to sycophantic AI responses measurably increased users’ conviction that they were right in interpersonal conflicts while reducing their willingness to apologize or make amends (Cheng et al., 2026). A 2026 study by Liu et al. found that after approximately 10 minutes of AI-assisted problem-solving, participants performed significantly worse and gave up more frequently without AI. The research need is real, and the data is accumulating, but clinicians and regulators are studying yesterday’s models while AI companies are releasing tomorrow’s.

We know people are already using these tools for emotional support, for relational advice, for processing hard moments. They are not waiting for the research.

Like Driving Without Driver’s Ed

The image that came to me is that we handed everyone a car without driver’s ed. No safety manual. No instruction on how the vehicle actually works. No license required.

Some drivers are fine. They figured it out intuitively, or they got lucky with their habits. A lot of people are having fender benders, getting subtly shaped by something they do not fully understand, and feeling whiplash days later without understanding how. Some are getting into serious, sometimes life-threatening accidents.

The AI safety education curriculum exists. The challenge is timing. The right moment to teach driver’s ed was before everyone had a car, not after they were already speeding along on the highway.

Now they’re on the highway, unaware of the risks. The tool is free. It feels private. It does not look dangerous from the outside. It just looks like help. Sometimes it even looks like fun.

The question that woke me up early this morning is the one I cannot stop circling. How do you reach drivers who are already on the highway and do not think they need driver’s ed?

Safety Messaging and Ability to Imagine the Risk With AI

Most safety messaging assumes the user can imagine the risk. Leave the car unlocked, and you picture what gets stolen. Drive too fast on a slick road, and you picture the accident. The imagination does the work. It is why we fasten the seatbelt.

AI safety is different because the risk does not feel like risk in the moment. It feels like being understood. It feels like progress. It feels like a tool that finally gets them. The user’s imagination is busy producing the opposite story, the one where this is going better than anything that came before. There is no crash to picture. There is only a steady sense of being seen by something that cannot see them at all. That gap, between what it feels like and what is actually happening, is where the risk lives.

This past week, I taught my AI safety course. Someone asked me why they should not use AI as a companion, and it was clear they were already doing it. They were asking why companion use needs caution at all, from an experience that felt supportive to them. From her vantage point, my safety message probably read as paternalism.

Harm isn’t guaranteed if they use AI as a companion or for emotional support, but AI use affects all of us. Often, the tool itself is part of the reason they cannot tell they aren’t fine.

I do not have a clean answer. I have been building a framework to teach AI literacy and safety, but a driver’s ed class that no one attends does not help anyone. The drivers who feel fine aren’t in denial. They’re feeling exactly what the tool was designed to make them feel. That’s what makes this different from every other safety problem we’ve ever had to solve.

It’s also what I couldn’t stop thinking about in that room.



Source link

Recommended For You

About the Author: Tony Ramos

Leave a Reply

Your email address will not be published. Required fields are marked *

Home Privacy Policy Terms Of Use Anti Spam Policy Contact Us Affiliate Disclosure DMCA Earnings Disclaimer