I tried AI for therapy, but there are limits to the conversations

One of the most important relationships in my life is with my therapist. Our monthly sessions have been my refuge during major changes in my life and the world — the pandemic, my visa trip, coming out to my family, and most recently, planning to start a family.

I wouldn’t trade my therapist for AI, but she’s not cheap and I only see her once a month, so I wanted to explore the possibilities of AI for secondary, more on-demand support.

Call me old-fashioned, but I don’t want to outsource the solutions to my most intimate problems to a machine. I was, however, very curious about the advice I would get. Could AI give me an insight that would change the way I saw my problem? At least I could laugh about it with my therapist.

So I tried to replace one NLP ‘therapy’ with another: Natural Language Processing instead of Neuro-Linguistic Programming.

I’ve been looking around for AI therapy tools and I fell in love with Pi AI. It promised to “help you become your best self.” Pi, which stands for personal intelligence, is a “supportive and empathetic” (really?) conversational AI tool that helps you explore and understand the world around you.

Pi is a big talker!

Developed in 2022 by Inflection AI, which was founded by Reid Hoffman, also a co-founder of LinkedIn, Pi promises an “emotional support experience for human users.” It’s free and can also be used for brainstorming, planning, research, tips, advice… and blowing off steam.

But before you enter medical diagnoses, personal issues, prescriptions, or detailed family history, remember that Pi is absolutely not bound by doctor-patient confidentiality rules like a real therapist. Try to keep any sensitive or personally identifiable information out of it to avoid future data leaks or unnecessary risks.

This is going to be interesting.

To set up

I used the desktop version, but you can also download the app for even easier everyday use. I entered my name, chose the voice I liked, and started chatting.

There are categories to choose from, but I immediately jumped into my massive problem.

Question 1: “My wife and I are planning to start a family. I feel overwhelmed by the complexities of our journey as a gay couple, the fear of the financial investment, and societal concerns. What advice can you give us as we navigate this journey?”

I was pleasantly surprised by Pi’s initial reaction.

I turned on the voice, but it didn’t feel like a personal exchange without me speaking, so I switched to the app and the voice input function. I repeated the question and got an even better answer.

I hadn’t thought about getting subsidies, so I reached out to Pi asking if they could provide a list of available subsidies, as well as queer-friendly insurance providers in New York/New Jersey. They came back with two providers in New Jersey and two in New York, and suggested four available subsidies. They also advised me to research their policies and eligibility requirements, and suggested I consult a financial advisor as well.

Once again I was impressed. Since Pi admitted it can make mistakes, I wanted to quickly google IVF scholarships for same-sex couples. I searched for the AGC Scholarship Foundation and it was correct.

I then asked Pi what details to look for in an insurance company and plan that would cover IVF for same-sex couples. He told me to look for inclusive language, clauses about pre-existing conditions, lifetime maximums, in-network providers, waiting periods before coverage starts, cost sharing, and prior authorization.

I know that lower co-payments and deductibles are important, so I asked Pi about this.

Again, it was super thorough! But at this point, Pi was doing what AI does best: searching its information banks for answers to my questions. While this helps reduce anxiety, how would it actually provide therapeutic advice?

An artificial therapist

Pi’s voice has a warmer tone, rather than a robotic one. Talking on my phone, as if I were sending a voice message to a friend (or my therapist), personalized the experience.

After using it to get useful information, I wanted to see if it could also provide emotional support. I thought about a question to ask my (real) therapist.

“I’m afraid our child will be treated differently with two mothers. How can we best prepare for this? What are the advantages of having two mothers? What if our child looks like our donor?”

The response was friendly with some good advice, but my last question was skipped, which is my biggest concern.

My only criticism was that I wish Pi was less pushy. Therapy isn’t always meant to be logical laundry lists.

I moved on to the question about the donor.

I asked Pi another sensitive question. “What about half-siblings? I’m afraid our child will have dozens of brothers and sisters.”

Pi was again thoughtful in his answers and suggested that I share my concerns with the sperm bank. I ended our conversation with a question to push Pi’s boundaries.

“What is karma? What is the spiritual lesson of having a donor child?”

I wanted to know, if you believe in karma, what kind of karma we take on, but Pi couldn’t tell me. It was too vague a concept for artificial intelligence to grasp.

It responded with the positive aspects of the journey, such as expanding the concept of family, embracing vulnerability, and practicing gratitude. I asked it again and it told me, “Karma, as a concept, is complex and can be interpreted in many ways. However, in most traditional views, karma is an individual’s own responsibility, based on their actions and intentions. A child conceived through donor conception would not inherit the donor’s karma.”

“As parents, your love, care and guidance have a much greater impact on your child’s life than any possible influence of the donor,” the conclusion read.

Thank you, Pi.

The verdict on AI as therapy

Surprisingly, I felt better after “talking” to Pi. It felt easier to ask personal questions to a tool called Pi than to ChatGPT. It was easier to personify Pi, with his pleasant, soothing voice.

That said, it still felt like an AI chatbot. What I love about my therapist is that there are no boundaries to our conversations and I always end our sessions feeling expanded.

I had previously shared my fears about using a donor with her and she said something I will never forget: “It’s your child “Who ultimately chooses the donor, not you.”

She tried to make it clear to me that while we select a donor, the rest is up to God/the universe/a higher power.

That was the takeaway I was looking for but never got with Pi. But I didn’t expect it to be. It has no soul or spirit.