Source- Unsplash Tim Mossholder
A small word that appears — and disappears.
I use generative AI in front of an audience regularly.
In classes, presentations and meetings, I will sometimes type at an AI prompt to demonstrate something about how models work. And what I type usually starts with one word.
“Please…”
A comment made in my last presentation made me start thinking about why I do this. “I stopped using those words with AI a long time ago,” floated through the air. I assume the words were please, sorry and thank-you.
And yet, I still typed, “Please,” in front of my prompt.
Why? Is it habit?
I’ve spent all my life saying please when I request something. It was drilled into me at home by my father and mother. These words were necessary. They create the social “oil” that makes communication flow. Without them, my parents would say, we could still communicate a request, but we might not receive the same response.
At least that’s what I was told.
Research on whether politeness improves AI responses is surprisingly inconclusive. Some studies suggest openly rude prompts produce poorer results. Others find little difference between ordinary requests with or without polite language. There doesn’t appear to be a simple “please equals better answer” rule.
So that doesn’t really explain why I keep typing it.
I don’t ask my calculator politely to divide two numbers. I don’t say “please” to a search engine. So, somewhere between search box and an AI prompt, my habits change.
One study suggested people often use “please” when they anticipate resistance. That made me wonder whether I was doing the same thing. I’m not sure I am. I don’t expect a language model to argue with me. But I do notice that “please” appears most often before we’ve established a conversational rhythm.
That made me pay attention to something else.
The word appears most often at the beginning of a conversation.
“Could you please summarize this paper?”
“Please review this draft.”
But after twenty or thirty exchanges, it begins to disappear.
Not because the conversation has become rude.
Because it has become normal.
I stop saying, “Could you please build an outline?” and simply ask, “Could you build an outline of that?”
Sometimes it becomes even shorter.
“Expand section three.”
“Compare that with the earlier version.”
“Let’s explore that idea.”
When I noticed this, it struck me that the pattern wasn’t unique to AI.
I see it in my own conversations with other people.
When I first meet someone, I am careful with my language. We signal respect. As we talk, we establish a rhythm. Once that rhythm exists, the conversation becomes more efficient. We interrupt each other. We finish each other’s thoughts. We say, “What about page three?” instead of, “Would you mind terribly if we discussed page three?” That just sounds awkward.
The politeness hasn’t disappeared.
It has changed form.
The same thing seems to happen in long conversations with a language model.
If I ask it to do something substantial — write a letter, generate a presentation, analyze a paper — I still find myself writing, “Please.”
If we’re already in the middle of exploring an idea, I usually don’t.
By then we’re no longer starting a conversation.
We’re continuing one.
I don’t know whether that says anything about artificial intelligence.
I suspect it says more about us.
Language isn’t just how we transfer information. It is how we establish relationships, even temporary ones. Once those relationships are established, we naturally stop repeating the signals that created them in the first place.
Perhaps that’s all “please” has ever been.
Or perhaps not.
After the presentation, I realized I’d demonstrated techniques for changing generative AI behaviour for forty-five minutes.
The comment that stayed with me wasn’t about the models.
It was about one word.
Top comments (0)