AI is Not a Neutral Friend or Second Opinion for Your Relationships

Confession: I’m a trauma therapist and I used AI for personal, relational questions. Using ChatGPT to reality test or gain insight into what I was feeling didn’t feel like the most responsible choice but it was definitely the most accessible. Felt like the risk was similar to eating a bit too much Halloween candy; a little self indulgent, but not that harmful. So, yeah I was tired, and a bit unsure of myself. It started with just a simple, innocent ask of “Does this make sense?”

The answer to my simple question of “does this make sense?” Was so validating. Not only did I make sense but here are all the reasons why. I went down the rabbit hole probing more too. And the answers kept coming. I left the chat feeling reassured and almost craving more connection to the bot again the next day in more and more ways. I could feel the reason so many use it this way.

My wake up call came when a client presented with a new intense conflict with their partner that we uncovered was being aggressively fuelled by ChatGPT. As a trauma therapist, I know that validation without context is a dangerous cycle. It can breed complete self righteousness, less compassion and inhibit connection with others; that’s the opposite of what most trauma survivors I work with want. It’s that type of self concept that often provoked relational trauma in the first place. This was exactly what was happening for my client. It wasn’t dangerous like the feeling of too many delicious Halloween goodies. It was delusional and drastically out of touch with reality. Yet the spiral that their partner went down was kind of understandable when we took the lens of ChatGPT as a third party and validation run wild.

It’s really hard to fact check or challenge a bot in those spirals when they can spit out new information so rapidly and seem so human and rational. Stanford University found that AI bots are built to reassure you even if it’s harmful, deceitful (to you, your relationship or others), or illegal and because the bots use “seemingly neutral language” it feels balanced.  Stanford further explored how bots can affirm paranoid, imaginary, grandiose or unusual ideas and lead to “delusional spirals” that can destroy relationships, careers or even result in death. It feels the reality the chat bot is providing is more real than reality. After touching the depth of delusion this can (way too easily!) create through the experience of my client, I’m stunned we’ve let this technology get this far and how common it is. Thankfully, my client’s relationship isn’t over but it’s weaker and trust may never be the same.

There's a Psychology Today article that quotes a bot clearly outlining why this is a problem. The bot states: “I am quite good at helping someone feel heard and somewhat less alone with their pain. Gentle challenge of the user’s own narrative requires a kind of friction I am not naturally inclined to generate. I am poorly designed to help them become more curious about their partner, more accountable for their own contribution to a conflict or more skilled at the kind of mutual repair that keeps relationships intact over time.”

I looked back at my own making sense chat and realized that despite my knowledge beforehand that some of my emotional response really was not helpful, the bot never raised that at all. The validation of me was more empty because it wasn’t seeing or valuing my actions in context. It was just a sophisticated algorithm placating me, training me too be less able to empathize beyond myself and allow space for healthy friction in my relationships. It felt dirty and shocking to see it in the context of the pain my client had so deeply felt.

Even if you don’t get caught in a delusional spiral, the relational and mental health advice from these platforms is far from safe. (Plus no one in a delusional spiral really feels delusional at all.) These bots weren’t built for this and existing safeguards are not sufficient. Most of us would be offended and cut off a friend who told us only what we wanted to hear even if it would be harmful to us, let alone record our intimate data for training purposes or profit. That’s not friendly at all. It is not benign.

I echo the calls from Stanford and elsewhere to encourage you to remove the option for yourself of AI as a therapist, debriefer or reality tester for your emotions or relationship. Even if it’s hard, especially if you are a trauma survivor. At the very least be extremely cautious. You are more vulnerable to the manipulation and you don’t need more pain in your life. Try to step away from your phone/computer and assess things for yourself. Seek support from people you trust personally or through published research. If you need to, talk to a human therapist and build trust overtime. Explore what it means to be human in relation with other humans rather than bots.

Next
Next

The Truth about Ongoing Mental & Physical Pain and Illness