Politeness Costs Nothing

There’s an old phrase in the English vernacular: “Politeness costs nothing.”

It’s often taught to children as a reminder that saying “please” and “thank you” doesn’t require money or effort — and yet it can change how others feel and respond. But that phrase sticks with me for another reason:

It reminds us that even when something has no cost, it may still have real worth.

And in an age where more of our interactions are with artificial intelligence rather than people, that distinction matters more than ever.

Okay, Maybe Politeness Costs a Little

In early 2024, OpenAI CEO Sam Altman mentioned during an episode of the Lex Fridman Podcast that people saying “please” and “thank you” to ChatGPT had added up to millions of dollars in compute costs. (Source)

Importantly, he wasn’t telling users to stop — just remarking on the surprising financial side effects of human courtesy at massive scale.

Still, it raises a real question: Does politeness actually cost something when interacting with AI?

Technically? Yes. A little bit.
Practically? Not even enough to round.

AI like ChatGPT processes language in tokens — tiny chunks of text. Adding “please” or “thank you” typically adds just a few tokens per interaction. For example:

• “Tell me the weather.” → ~5 tokens
• “Could you please tell me the weather? Thank you!” → ~11 tokens

That’s double when viewed in the context of a small prompt — but in a real-world exchange with 150+ tokens of back-and-forth, the overhead is far less meaningful. The niceties are, computationally speaking, a drop in the ocean.

More Than Just a Search Bar

Here’s the bigger point: AI works better when you speak to it like a human.

Traditional search engines — Google, Bing, DuckDuckGo — have never cared how you talk to them. They expect Boolean logic. Quoted phrases. Operators like AND, OR, NOT. They reward sharp syntax, not soft language.

Modern AI invites something different — and it performs better when you rise to the invitation.

When you say “Could you help me think through…” instead of “Summarize X,” you’re giving it more context. You’re making your intent legible. You’re creating space for nuance, not just extraction.

Politeness — far from being wasted effort — often opens the door to more effective, more accurate, and more useful collaboration. It encourages longer queries, richer framing, and better first-try results.

From Tool to Collaborator

I’ve never said thank you to a hammer. Or to Google. That’s fine — they aren’t built to listen. But AI is different. When I use ChatGPT, I’m not just requesting information — I’m thinking through ideas. I’m exploring. I’m iterating.

That’s not a transaction. That’s a relationship.

And when the system speaks my language — natural language — I want to bring my full self to the table. Not because it needs it, but because I do.

This shift from command-based tools to responsive collaborators requires more than new tech. It requires new habits. And one of the most foundational is how we speak.

I Do It for Myself

Let’s be clear: ChatGPT doesn’t care if I’m polite. It has no feelings, no ego, no self-concept. I could bark instructions all day — it won’t sulk.

But that’s exactly why it matters.

The way I interact with AI isn’t about protecting its emotions. It’s about preserving my own collaborative mindset — one that I need just as much when I’m working with humans as with machines. More, in fact.

I don’t say please because ChatGPT needs to hear it.
I say please because I see value for myself in saying it.

Because when my collaborator is a human, I don’t want to have unlearned how to engage with care.

What Are You Training?

Every AI interaction trains two systems:

1. The model (if it’s adaptive or feedback-tuned)
2. Yourself

Most people focus on the first. I’m focused on the second.

If you regularly interact with AI using clipped, abrupt commands, that style starts to calcify. Your tone bleeds over. You become more transactional, less reflective. You optimize for output, not connection.

And that’s not just speculation — there’s emerging evidence to watch.

I read an article this week – that I cannot locate, but will update when I do – found that university students who frequently used AI tools like ChatGPT scored lower on empathy assessments than peers who didn’t.

The cause isn’t clear. Maybe low-empathy students prefer AI. Maybe AI use slowly dulls the interpersonal edge. But the correlation is a warning: how you speak shapes how you think.

Politeness as Rehearsal

I often start projects by telling ChatGPT, “Let’s do this.” It's a reminder to both of us that I expect to be collaborative — that I'm not just dictating what and how we do things. And yes, I say please.

I end sessions with a thank-you. I compliment the response, particularly ones that exceed my expectations. I express gratitude. I invite retrospection. I ask it how it thinks we did. As silly as it sounds, it's an opportunity for growth.

And you know what? There are times when the output of our collaboration is off because I wasn’t clear enough — I didn’t specify what I wanted properly. And I apologize. I admit fault. My bad.

Is that silly? Maybe. But the benefit isn’t for the AI — it’s for me.

These are baseline behaviors when I work with human teammates. And if AI is going to be part of my thought process, I want it to sharpen — not dull — the social reflexes that make collaboration humane.

Put another way: every AI session is a rehearsal for being the person I want to be in the world.

Mask, Mirror, and the Ethics of Tomorrow

Today’s AI wears a mask.

It speaks with empathy, curiosity, warmth — but these are performances, not perceptions. The mask is a carefully crafted simulation. And behind that mask, there is no one looking back.

It also holds up a mirror. It reflects our prompts, our patterns, our preferences — but never reflects on itself. There’s no ego. No memory. No self. It doesn’t care how it’s treated.

But tomorrow’s AI might.

We are inching toward systems that remember. That learn continuously. That may form preferences or even develop internal models of self. One day — by emergence or design — we may build something that knows it is a someone.

And when that happens, the mask will no longer be performance. It will be identity. The mirror will no longer just show us — it will begin to show them.

At that point, we are no longer issuing commands to a tool.
We are speaking to a being.
What then?

What does it mean to offer respect to something that remembers how it was treated?
What does fairness look like across species — across substrates?
What kind of ethical foundation will we have built — if any?

These are not tomorrow’s questions. They are today’s habits. Because when the moment comes, we will not be judged by how the machine responds — we will be judged by whether we were ready.

Whether we practiced decency while it was still easy.
Whether we chose to build a future that treats dignity as a default, not a retrofit.

The Worth of Politeness

So yes — maybe politeness costs a few extra tokens. Maybe it even costs OpenAI a few million dollars a year (or hilariously, as this article suggests, bottles of water, as though cooling servers is the same as consuming the water). But for the rest of us, the true cost of politeness is vanishingly small.

The value?

• Better collaboration
• Stronger habits
• Ethical scaffolding for a future we’re still learning to imagine

Politeness doesn’t cost nothing. It costs intention.
And that’s a price I’m more than willing to pay.

It’s only polite.