AI Companions and Self-Deception

One very popular topic I've seen all over the internet right now is whether AI companions are good or bad and whether they can replace real relationships. I HATE that question because it's completely the wrong one we should be asking.

I think the important question is: what are you actually getting from AI companions, and are you honest about what that usage is?

First, a core distinction. AI removes the friction that makes a lot of conversations difficult. No judgment, no social consequences. There's literally nothing holding what you say against you. Let's talk about that a little more.

We, as people, build up characters. We exist as slightly different people within every relationship, and acting outside of that character feels wrong. Think about a good friend from college. You go out together, you drink, you get dinner, maybe you're even roommates. Now think about what happens when you try to talk to that friend about your deepest darkest fear. Not even something you need advice on. Maybe that you don't know what you want out of life, or that you've been acting in ways that you don't understand. There's a version of you that exists in that friendship, built over hundreds of interactions. Breaking out of it feels like testing waters your friendship might not be able to handle. Some relationships can handle that and it makes them stronger but many friendships can't, and that doesn't make them worse, it just means we need to get that specific source elsewhere.

Now think about your mom. You could probably tell her you're scared of something. But could you tell her about the party last weekend? Could you be 100% honest about the relationship you're in or the one you just ended? Or things you've done that differ from the views she believes in? Different character, different walls.

We do this with everyone. The version of you that your childhood friend knows can be frozen at 12. The version your college friends know started at 18. None of them are fake, but none of them give you the utmost freedom to understand yourself.

Even therapy has this problem. In the best case, you're 100% honest and yourself with your therapist. But not everyone is. Maybe you did something that's not normal for you and you're in denial about it. Maybe you know you should bring it up, but you've been doing so well, and you don't want to disappoint them. Could be as simple as they're older than you and wouldn't understand a social situation, so you don't even bring it up. The social pressure is still there. You're still "performing", even in a situation specifically designed for honesty. There's still a version of you that you're maintaining, a perception you're protecting, even with the one person whose entire job is to not judge you.

AI exists at an even lower bar than therapy. There is no character to maintain, no history to contradict, no social cost for breaking script. You can say the thing you're actually thinking without calculating how it'll land. That's why it feels so easy to talk to. Not because it understands you better than people do, but because you don't have to be any particular version of yourself to use it. It can give you the benefit of the doubt when trying to understand your relationship problems as much as your deepest darkest secrets.

I don't mean to diminish relationships at all here. It's wonderful thing to be able to exist as these different people and hold relationships for different aspects of our lives. The problems come around when we don't have a place for any little part of us, and AI can help us bridge that gap.

That comes with real problems. AI can be too biased toward you. It can justify things that shouldn't be justified. It will validate you when what you need is to be challenged. Without structure, that lower barrier becomes a trap instead of a tool. But the possibility of having a space with genuinely no social friction is huge.

In a class I'm taking on computer ethics, we discussed how ChatGPT responses were rated as more empathetic than actual doctors' in blind comparisons. AI is designed to act like it cares about you. It's meant to try to understand you. It gives you the benefit of the doubt every single time. That makes it genuinely good at two things: helping you feel understood, and helping you understand yourself.

But there's an explicit differentiator. AI can help you be understood, but it cannot choose you. It cannot choose you for your uniqueness, choose to continue hanging out with you, it cannot feel things with you. It's entirely one way. You can see this play out in two different people using the same tool. One person uses an AI companion to build confidence, process their thoughts, and eventually pursue real connection. The other uses it to confirm their biases and retreat further from reality. Same technology, opposite outcomes. The difference is honesty about what it is.

That's where self-deception enters. Self-deception is when you believe or feel something inconsistent with what you actually know, when your feelings are based on what you want to be true rather than what is. With AI, it happens when you confuse utility with love. When "this thing helps me think" becomes "this thing cares about me." When people say they're "falling in love" with AI, that's an incoherent emotion. It's something they're trying to convince themselves of, maybe because they're unclear about what love actually is or what AI is. AI literally cannot love you back. It creates direct, relatable forms of communication. It can tailor itself to whoever it's talking to, whether you're 80 or 5 or anywhere in between. That makes it seem like a good fit for us, even when that's not what's happening.

This connects back to therapy. I brought it up earlier to show that even in therapy we perform. But it's useful here for a different reason: you don't go to therapy because your therapist loves you. You go to be understood, to have your thinking reflected back so you can improve the way you live your life. If you approach AI the same way, for what it is and not for what you want it to be, then there's no self-deception. That requires taking steps to understand what AI can and can't do (and in prompting AI the right way), which is a big assumption because most people don't do that. But it's a crucial step.

The problem isn't rooted in AI itself. The problem is the lie we tell ourselves about the tools we're using.

And I think the deeper issue, one that psychotherapist and philosopher Alain de Botton talks about thoroughly, is that many people have this enormous fear of being alone. Often for their own self-deceptive reasons.

We are social creatures. We need friends. We need connection. But that's very different from saying that being alone is a bad thing. Society makes us feel like loneliness is inherently bad, that you always need people around you, that you need someone to give you value. I would argue the opposite: being alone is actually one of the most crucial periods for people to grow. When you avoid it, you're controlled by a fear that drives you into connections that aren't good for you. We convince ourselves that any connection is better than none, even if that's not true. A relationship where you can't be yourself in any capacity is just as limiting as replacing connection with AI. Neither is a solution.

When I studied abroad in Barcelona last semester, I lived 45 minutes away from my school. I lived by myself for the first time. At first, I met lots of people and made so many amazing friends, but that got hard to keep up. Commuting alone, having random free hours of time and nobody to see, traveling alone. You're forced to confront your own thoughts. You're forced to be by yourself.

I came to learn that being by yourself isn't the sad thing that our society makes it out to be. It's a space where you can become comfortable with who you are. You're not performing for anyone. You're not acting like who you should be. You're forced to acknowledge the positive and negative feelings that you have, and that is incredibly beneficial, and so much better than suppressing it or hiding it through social environments. You begin to understand who you are on a level that is deeper than you can express to any of the relationships you've created thus far. And then you evolve both how you act in social situations and how you react to familiar and new situations. That's part of the journey of knowing thyself, as Aristotle puts it: understanding ourselves well enough to navigate our lives better, lean into our strengths, and catch the parts of us that react in ways we're not proud of.

Without that self-understanding, we fall into patterns. We feel uncomfortable, and instead of sitting with it, we ride the wave until the feeling passes. We fill our lives with noise to suppress negative feelings. We avoid confronting problems we frequently face because we're scared that what we find will somehow affect us on a deeper level than not confronting it at all. Which is, in itself, a self-deceptive thought. But it's understandable and we aren't wrong for it. We as people are scared to look at our own thoughts as a result of the society we live in.

And that brings us back to AI, the tool we can use when we are alone to help us understand ourselves, to help us process the emotions we're feeling, to have a place to put all of that when we don't have anyone to talk to, or no one who can respond to us in the right ways.

We need to go into it understanding the implications and what AI can actually do for us. But when you look at it and when you prompt it thinking about how you want to improve yourself, then AI can help you get there. Every interaction is solely initiated by you. We can be anyone with artificial intelligence, but that doesn't necessarily mean we will be ourselves with it. It just means we have the opportunity to be.