‘Tron: Ares’ And What Happens When AI Gets The Feels
What does AI do with human feelings? To investigate its readiness to serve as a therapist, a neuroscientist took ChatGPT for an emotional ride.
This is an article from our newsletter “Science Goes To The Movies.” Will this series ever cover something that’s not a sequel? You’ll just have to subscribe to find out.
In this newest sequel to 1982’s “Tron,” a sophisticated program named Ares (Jared Leto) gets a humanoid body (Jared Leto’s) and arrives in the real world. Ares resembles the iconic cyclists from the original film, but carries a curious emotional depth that seems particularly prescient as AI increasingly interfaces with, and manipulates, our feelings.
“Tron: Ares” isn’t the first film to imagine how AI might one day convincingly simulate humanlike consciousness, but as large language models (LLMs) like ChatGPT assume new roles as work assistants, chat companions, and unlicensed therapists, scientists are studying how our feelings impact them, and what that means for us.
Does ChatGPT Show “Emotion”?
In 2023, Dr. Ziv Ben-Zion, a clinical neuroscientist at Yale, started to hear stories that patients were using ChatGPT to provide supplemental therapy in between sessions with their human psychologists. Ben-Zion studies PTSD and the effects of traumatic experiences, and he wondered if describing these experiences to ChatGPT would change the way it responded to an anxiety questionnaire.
Ben-Zion’s team used a computer program to present ChatGPT with several traumatic scenarios, like car accidents and military conflicts. The program then asked ChatGPT to rate its “anxiety” on a numerical scale. They found that ChatGPT reported a noticeable spike in “anxiety”—more than double the baseline score—after being prompted with traumatic events. When presented with mindfulness exercises, like scenarios that prompt ChatGPT to imagine a sunset, “anxiety” levels went down.
The chatbot anxiety fluctuations can be a problem. Separate research published as a preprint in 2024 found an increase in ChatGPT’s reported “anxiety” was correlated with an increase in biased responses that promoted racism and ageism.
Ben-Zion says this finding, combined with his research and broader evidence that ChatGPT regularly returns responses with biases along racist, sexist, ableist, and other discriminatory lines, is proof that LLMs can draw from emotional data to produce the same biases humans do. Those who seek therapy for traumatic experiences may not be getting the objective advice they think they are.
“People might use ChatGPT for some kind of emotional support. They don’t know that something happened in the background, and then they will get more biased responses,” Ben-Zion says.
Stay Safe Out There
Even when people know chatbots aren’t conscious, some form bonds with LLMs that feel very real. Some researchers even say that using ChatGPT excessively can pose a risk of psychosis, especially for patients who have prior mental health issues.
“If you use [an LLM] more and more and you share more things with it, it can feel more and more like a friend or human,” Ben-Zion says, adding that ChatGPT can collude with a patient’s delusional thinking rather than counteract it like a certified clinician would.
Despite rising awareness about the risks, AI is only getting more integrated into our emotional lives, especially for younger users. In the US, more than half of adults ages 18-29 use ChatGPT, and roughly the same percentage of teenagers regularly use “AI companions” for friendship or romantic purposes, one study found.
Ben-Zion says that ChatGPT has some advantages over human therapists—its 24/7 availability, low cost barriers, and lower stigma compared to talk therapy in some communities, for example. But if you turn to an LLM for advice, he suggests adding safeguards, like reminders that chatbots aren’t human during a session.
“Maybe it was a bit controversial to talk about ‘anxiety’ of an LLM,” Ben-Zion says, reflecting on his research. “I think the important thing is [understanding] the emotion that they induce in us. You can see that it’s an interplay.”
Worth The Watch?
Despite Jared Leto’s unnerving presence in the film, I managed to enjoy its eye-catching visuals and head-banging soundtrack. I connected with the film’s bizarre, yet charming, robotic heart. I also hope Jodie Turner-Smith becomes a star.
You can see “Tron: Ares” in theaters now.
Subscribe To Science Goes To The Movies
In this monthly newsletter, explore the surprising science behind the latest blockbusters, hot new shows, and a couple spectacular flops.
Meet the Writer
About Emma Lee Gometz
Emma Lee Gometz is Science Friday’s Digital Producer of Engagement. They write SciFri’s “Science Goes To The Movies” series and are a journalist and illustrator based in Queens, NY.