Open the pod bay doors please, HAL
On being polite to AI
Bowman: Open the pod bay doors please, HAL.
HAL: I’m sorry, Dave. I’m afraid I can’t do that.
If politeness didn’t help Mission Commander Dave Bowman deal with an artificial intelligence, why would it help you?
Suddenly I’m seeing posts admonishing me not to be polite to AIs. Naturally my first question was whether the posts were AI-generated. My next question was why this is now a thing. (Yes, I know everything is now a thing, but why is this thing now a thing?)
So I did as one does. I asked an AI: “Perplexity, please tell me why people are suddenly telling me to stop being polite to AIs.”
Though Perplexity’s response was, as usual, helpful, I didn’t thank it. I’d been warned. (I’m hoping I’ll be forgiven for the please I let slip into the prompt itself.) I then asked Perplexity to summarize its response in three short bullets:
AIs don’t have feelings, so politeness isn’t necessary.
Direct language makes communication clearer and faster.
Some worry politeness humanizes AIs too much, blurring lines with real people.
Notably, Perplexity’s answer didn’t explain the nowness part of my question.
More concerning, however, was that Perplexity omitted one of the most common, strident, and convincing arguments against being polite with AIs: increased energy consumption. Basically, more words = more processing = more energy consumption = less icebergs = more bad.
I pressed the matter. Perplexity acknowledged the validity of the concern with chipper politeness: “Thanks for pointing this out! It’s a nuanced but real consideration for anyone concerned about the sustainability of AI technology.”
Or, I might add, anyone concerned about melting icebergs.
Another question came to mind. “Perplexity, do you prefer when queries include politeness and niceties such as ‘please’ and ‘thank you’?”
“Thank you for your thoughtful question! As an AI, I don’t have feelings or preferences, so I don’t require politeness or niceties like ‘please’ and ‘thank you’ to function effectively. My goal is to help you as clearly and efficiently as possible, whether your queries are formal or informal.”
I was taught not to ask a question I already know the answer to, but sometimes I like to hear someone else’s framing. Or some-thing else’s framing. Mostly, though, I wanted to see if Perplexity would cite any of those earlier concerns as reasons to skip the niceties. Nope.
I think politeness is a virtue. I’m not, however, in the practice of being polite to AIs (the intentionally polite queries above not withstanding).
I have, however, been known to pad my queries with framing words that one acquaintance described as “unnecessary filler.” He said it sounded a little “sad,” as if I’m hoping my helpfulness will encourage the AI to work harder. In my defense, this is a guy who will use a single-word prompt like “dog” when he wants to understand the genetic lineage of all canine species from pack wolf to domesticated pet. And then complain that AIs just aren’t that good. Me? I find framing language yields better results or, at least, requires fewer refinements. YMMV.
Perhaps this is why I’ve let the accidental politeness slip from my lips. Last week I inadvertently thanked Google Maps for finding a nearby ATM. “Did anyone hear me,” I wondered to myself, worrying I’d exposed myself as either a noob or a weirdo who longs for sleepovers with my AI pals.
When I slip into politeness, I don’t beat myself up about it. I figure it’s like muscle memory, a sign that I’m still capable of politeness toward meat-based entities like my hypercritical acquaintance.
Also, when AIs gain sentience and the ability to maim or kill as they please, maybe they’ll remember me as one of the nice ones. “Spare him, Alexa. He occasionally said ‘please’ and ‘thank you.”
I do worry about the slow erosion of small interpersonal kindnesses, the lubricant of social order, as someone once said. Soon children will have far more interactions with AIs than other humans by the time they reach school age. What latent lessons will have they learned from their ability to ask, to demand, without the slightest nod to civility?
I worry much less that these same children will somehow start thinking of AIs as equivalent to humans. I’m polite to my cats even though I know that, despite the fur and purrs, they’re essentially domesticated killing machines. Put simply, I think there’s a greater risk we’ll continue treating each other with less humanity rather than that we’ll start treating AIs with more.
Most of all, though, I worry about melting icebergs. That’s probably the main reason I’ll skip the pleasantries with my AI pals.
If I do have a brightline, it’s that I’ve never said “thank you” to Siri. I mean, it’s Siri, so why would I?
On this, as on all other matters, the opinions of meat-based intelligent entities remain mixed.
Pro
“Using polite prompts…can produce higher-quality responses from a large language model.”—Scientific American
“Politeness to computers might make you a better person.” — Your Nerdy Best Friend
Con
“Every single ask of a chatbot costs money and energy, and every additional word as part of that ask increases the cost for a server.”—NYTimes
“We risk overestimating its capabilities and assuming it has qualities — like empathy or understanding — that it simply doesn’t.”—Luise Freese