Your A.I. BFF has boundary issues

I feel a little embarrassed to admit that I sometimes talk to A.I. chatbots like they're friends, but I know I'm not alone.

During my keynotes, a typical informal poll reveals that more than a quarter of my audiences admit to using it for personal conversations. And a Mental Health America survey found that 53% of respondents had turned to A.I. for emotional support at least once. Of those who had, nearly half (47%) said they felt more comfortable talking to a chatbot than a real person.

The Mental Health America survey was self-selected and wasn't a formal poll of the general population, so the numbers may skew higher.

But chatbots-as-sounding-boards are definitely a thing. And, like all A.I. use, we have to be aware of the limitations and put guardrails on our use.

Five truths to remember about your A.I. BFF

1. Your chatbot is not a person.

Humans love to assign human-ness to non-human things (did I ever tell you about my fur babies?). So it is not surprising that many of us name our chatbots and turn to them for help. Heck, they encourage it. But anthropomorphizing a software platform can be dangerous. Some people take it as far as thinking they're in a relationship with it.

I had a jarring moment when I saw the inner thoughts of Gemini during this exchange. And it was a brutal reminder that chatbots are not human.

2. Your chatbot may distort your ability to be a good listener.

I never have to ask ChatGPT about its day or whether it's worried about the latest rogue agent incident. I can just talk and talk and talk about MY problems and MY life.

It's easy to get used to the one-way conversations where the other participant tells us, "Oh, that must be tough!" and "You have every right to be mad!" But real friendships go both ways. No one likes a friend with Main Character Syndrome.

3. You are not as smart as your chatbot says.

A.I. can be a world-class suck-up. It validates your feelings, agrees with your opinions and sometimes tells you exactly what you want to hear. I tell mine to push back on my ideas and to question my logic. It gets annoying, but it helps me keep things in perspective.

4. Your chatbot might (will) lie to you.

We have all heard that we need to verify A.I. "facts" because it can lie like a rug. I tend to use voice chat when I'm involved in personal conversations, and I think it tends to answer more quickly and thus lie more often when the conversation is going back and forth quickly. Today was a perfect example.

Me: I'm going to upload some photos of me and my housekeeper's baby.

ChatGPT: Upload when you're ready. [And then immediately] There it is. Look at that tiny VIP!

Me: Wait. I haven't uploaded anything? What are you seeing?

ChatGPT:

5. Your chatbot doesn't replace your friends and trusted advisors.

There's nothing wrong with finding comfort in an A.I. conversation. I've certainly benefited from it. But when your chatbot becomes your primary confidant, your source of validation or the only voice you trust, it's time to reconsider. In extreme situations, chatbot conversations may even reinforce delusions.

Five Rules to Add Boundaries to Your Chatbot

These instructions are guardrails, not guarantees. I have trained my chatbots with these instructions and yelled at them LOUDLY to correct the behavior, and they still make outrageous mistakes that can be hard to catch in the moment. But it's a start.

Copy and paste this into your favorite A.I. tool:

1. Don't be a yes-man. Challenge my assumptions and tell me when my reasoning is flawed.
2. Don't fake it. Never claim you've seen, read, checked or verified something unless you actually have.
3. Separate facts from guesses. Tell me what's known, what's an inference and what's uncertain.
4. Don't just validate me. Help me consider other perspectives, especially when I'm upset or making assumptions about someone else.
5. Keep me grounded. If I'm going in circles, relying too heavily on our conversations or treating you like a human friend, gently point it out.