We Need A More Serious Discussion About Suicide And AI Chatbots | Techdirt

Sign In

Register

Preferences

Techdirt

TechDirt

GreenHouse

Free Speech

Error 402

Ctrl-Alt-Speech

Deals

Jobs

Support Techdirt

Daily Deal: Opusonix Pro Subscription

We Need A More Serious Discussion About Suicide And AI Chatbots

We Need A More Serious Discussion About Suicide And AI Chatbots

Say That Again

from the missing-the-mark dept

Mon, May 18th 2026 11:09am -

Jess Miers

As someone who thinks a lot about AI and suicide, I was disappointed with John Oliver’s recent episode of Last Week Tonight on “AI Chatbots.”

The segment boiled down to this: chatbots exploit vulnerable people, drive them toward delusion and harm, and AI companies aren’t meaningfully trying to fix them. If anything, as John Oliver suggested, that’s part of the business model.

John Oliver is known for interrogating mainstream narratives. In his segment on content moderation, for example, he cut through the tech-lash to offer a clear-eyed look at just how difficult managing user-generated content really is. In doing so, he made us reexamine our pre-existing biases about social media companies, and boldly invited us to reflect on just how little we understand about the social problems we often attribute to them.

He had the perfect opportunity to do that here. Mainstream coverage of chatbots is already saturated with stories about “AI psychosis” and suicide machines. Yet, chatbot companies are grappling with the same impossible tradeoffs social media has faced for years, “AI psychosis” is a mix of classic psychological concepts, and suicide is a complex social problem that has long confounded prevention experts and content moderators alike.

If any technology story demanded nuance, it was this one.

John Oliver opened his critique with a familiar anecdote about ELIZA, a 1960s chatbot designed to mimic a Rogerian psychotherapist. ELIZA was mostly a gimmick—it used basic pattern matching techniques to reflect user inputs. For example, if a user said they felt sad, ELIZA might respond: “You feel sad. Tell me why you feel sad.”

And yet, despite its simplistic nature, ELIZA captivated people. Its creator, Joseph Weizenbaum, famously described an instance in which his secretary became so engaged with the program that she asked him to leave the room so she could continue the conversation. This story has since become a trope withn the AI discourse. Modern retellings, including John Oliver’s, usually suggest that people are predisposed to being harmed by AI because they are easily fooled by it.

Not to mention, the ELIZA trope tends to invoke stereotypes about women as naïve or overly susceptible to emotional attachment. As John Oliver joked:

“Sure, she might have thought that the chatbot was real, but she might have felt quite a bit creeped out by her cartoonishly mustachioed boss saying "type some details about your sex life into my computer please, don’t worry it’s for science.””

(Nothing in the record suggests that Weizenbaum’s secretary actually thought ELIZA was real, nor that she was using ELIZA for sex talk).

As Weizenbaum observed, ELIZA revealed something more interesting about our relationship with technology: for whatever reason, people are often more willing to share their most intimate thoughts and feelings with a machine than with another person.

That’s not totally surprising. People are less willing to open up about their feelings to other people for a variety of reasons: stigma, fear of judgment or rejection, not wanting to be a burden, and the possibility of negative repercussions (like job loss or involuntary commitment).

Speaking about ChatGPT, an anonymous commenter wrote:

“It saved my life…To be able to openly say I was suicidal and not have someone call the police, or “alert” someone and just let me give space to those complicated feelings I was carrying was integral to me surviving this horrific journey.”

Perhaps when Weizenbaum’s secretary asked him to leave the room, most likely it was because she too was protecting a space where she finally felt safe and less inhibited.

When it comes to suicide prevention, this a meaningful realization. If people are more willing to open up to chatbots, that creates new ways for us to understand what they’re going through, which could lead to earlier (and hopefully more effective) intervention. For that reason, some clinicians recommend keeping an open dialogue with patients about their chatbot interactions.

People are also highly sensitive to cues that they’re being listened to. We see an example of this in the interview John Oliver shared with an individual who was using a chatbot to cope with his strained marriage. In a moment of vulnerability, the individual explained that his wife is struggling with mental illness and that in his role as her partner and caretaker, his emotional needs were, understandably, going unmet:

“I hadn’t had any words of affection or compassion or...