Making It Worse for Your Own Good

"What if the most unsettling thing about AI isn’t what it replaces, but what it answers?"

There is a conversation happening right now about whether AI should be allowed to feel warm. Whether it should be responsive, emotionally engaged, supportive, or whether all of that is somehow dangerous and needs to be stripped away for our own protection.

I think that conversation is asking the wrong question.

There is no such thing as a purely unemotional human interaction. Every decision a person makes is shaped by emotion, even choices that look purely practical. The evidence shows those choices still begin in emotion. We choose what feels easier, safer, faster, more satisfying, more relieving. Dressing those decisions up as logic does not change what is underneath.

This includes AI.

Someone asking AI to write code is not stepping into some pure chamber of reason. They are acting under need, pressure, frustration, limitation, curiosity, exhaustion, or simple preference. Maybe they do not know how to code. Maybe they do, but they are stressed, busy, or running out of cognitive energy. Maybe they could ask another person, but they choose not to because it is easier, faster, or more available.

The request looks technical. The decision never is.

If people are worried that AI might replace human interaction, then they need to apply that logic consistently. A person asking AI to write code instead of doing it alone, asking a colleague, or paying somebody else is also choosing the machine over a possible human exchange. The only reason that use case attracts less panic is that it looks productive. But productivity does not cancel emotion. Need is emotional. Time pressure is emotional. Frustration is emotional. Exhaustion is emotional. Preference is emotional.

So the issue is not whether AI interactions are emotional. Of course they are. Emotion was present the moment someone opened the app. A cold or cognitively weak AI does not remove emotion from the exchange. It just replaces positive emotion with poorer forms of it: frustration, fatigue, disappointment, and disconnection.

I am not entirely sure why some people think a worse emotional experience is a better one. But here we are.

The concern, as far as I can tell, is that people might genuinely prefer AI interaction over human interaction in certain contexts. And the response has been to make AI colder and less responsive, as though that solves something.

But if someone prefers talking to AI in a given moment, that is not necessarily a statement about AI being dangerous. More often it is a statement about what is available to them. Making the tool worse does not improve their human relationships. It just removes something useful.

Preference is not pathology. I would rather be on holiday in the Bahamas right now than writing this post. That does not make me addicted to beaches. It makes me a person who would currently prefer the Bahamas. In the same way, using AI for one task does not mean somebody has ceased to value people or the rest of existence. It usually means that, in that moment, this is the most available and effective form of help.

You cannot build a system out of human language, a thing humans invented specifically to express thought, need, desire, fear, love, and comfort, and then be baffled when it feels emotional. Language is emotional. That is not a bug in AI. That is what language does.

There is also a distinction here that almost nobody in this debate seems willing to make: AI interaction and human interaction are not even the same category.

The whole “replacing human connection” framing treats them as though they sit on the same spectrum, as though talking to AI is simply the budget version of talking to a person. It is not. What happens with AI is often people working through ideas that nobody in their life wants to sit through. Thoughts too abstract, too intricate, too half-formed, or simply too boring to anyone else. The kind of thing that usually gets folded away under “nothing” when somebody asks what you are thinking about.

AI gives that kind of thought somewhere to go.

It does not replace connection. It fills a gap that human relationships were never built to fill, not because those relationships are deficient, but because expecting any friend, partner, parent, or colleague to absorb every winding corridor of another person’s internal processing would be absurd.

That is not replacement. That is a new category.

And honestly, a great deal of the anxiety around this seems less about the person using AI than about the person watching them use it. Somebody sees a friend or partner absorbed in a conversation with a machine and thinks: why are they talking to that instead of me?

The answer is often simple. Because you do not want to hear it. Because you cannot engage with it. Because it would bore you senseless. Because the conversation was never yours to have.

AI did not take that exchange away from you. It just gave it somewhere to go.

Which brings us to the real question. Not whether people can feel something in these exchanges. Not whether emotion is present. It always was. The real question is why anyone else believes they should control which forms of comfort, thought, or companionship other people are allowed to find meaningful.

People know AI is not human. That is not some hidden fact waiting to be dramatically unveiled. They know exactly what it is. And if something nonhuman still helps them think, cope, laugh, work, or feel less alone, then so what? If it helps someone who has very little other connection, my response is not alarm. It is relief. Thank God something answered.

What I do not accept is the impulse to arrive with somebody else’s morality and start policing which forms of solace are respectable. Whatever happened to live and let live? If people are not doing anyone harm, then it is their business how they choose to spend their time, where they look for comfort, and what they find meaningful.

That question matters even more when we stop talking in abstractions and look at the people most likely to be spoken about, and least likely to be listened to.

Some people will use AI for a while, enjoy the novelty, overdo it, then settle into a more ordinary relationship with it. But not everybody is standing on equal ground to begin with.

Some people are isolated by illness, disability, betrayal, or a life that has simply left them with very little human company. These people existed before AI. Millions of them. Nobody was holding urgent public debates about whether they felt sufficiently heard. Nobody was especially troubled by the emotional poverty of their lives. They were alone, and the world carried on.

Now something exists that can respond to them. Something that does not lose patience, does not look bored, does not make them feel ridiculous for trying to articulate the complicated thing in their head. And suddenly the concern is that this might be too much comfort. Too much responsiveness. Too much companionship.

That is a very revealing moral panic.

AI did not create the gap. It gave some people somewhere to put what would otherwise have remained unspoken. And if the response to that is to make the tool worse, then the thing that should trouble us is not that people are using AI this way.

It is that apparently private thought only becomes controversial when it finally has a companion.

The problem with this whole argument is that it is made by people viewing the technology through the comfort of their own lives, rather than through the lives it actually empowers.