All Robots And Computers Must Shut The Hell Up

The language our machines use to speak to us has changed.

We wanted to humanize the computer. We wanted it to meet us where we are, with our evolved social primate brains, and simple declarative statements like "File Not Found" are just a tiny sliver of what we evolved social primates actually use language for. Language embeds a constant flow of metadata regarding the speaker's intent, attitudes, emotional states and what have you. Micro-affirmations, SYN-ACK handshakes, "yes I am listening", "I am reporting my internal state as concerned". Sometimes the metadata is the significant data and the literal words are just a carrier. "Oops" has no referent beyond the speaker conveying an attitude of casual apology.

The "classical" dialect of software messages is, in contrast, strictly literal and informative, and therefore less "human". Someone who's just banged a shin on the coffee table will demonstrate some prodigious vocabulary use, but you won't hear anything like Error 384: Tiba Integrity Violation . Therefore (goes the argument for humanizing the computer), computers should communicate messages beyond the literal and informative - messages indicating success should also convey elation ("hooray! The update completed!"); failure messages should be appropriately contrite ("oops, we can't reach the network"). The user is no longer operating a complex, often cryptic machine, but speaking with a companion.

The flaw in this reasoning lies in the initial premise - we should not humanize computers, because computers are not human. Computers are tools. If you stick googly eyes on a hammer you've just made a worse hammer.

Sometimes people will rag on Linux with an anecdote of trying to do something simple like fetch OS updates and getting an arcane, incomprehensible error message like Could not get lock /var/lib/apt/lists/lock - open (11: Resource temporarily unavailable)¹ But the thing about "classical" error messages like this one is if you read them you will learn what the problem is. Or if it uses terms that go over your head you can Google it to learn what the problem is in more basic language, and you also gain some knowledge about how your computer works. Maybe you can even fix it without googling next time!

The "friendly" Windows equivalent message would be something like "We couldn't get updates right now, try again later". Does that "humanize" the user interaction? Maybe, but it also withholds information and makes troubleshooting much harder. The signal-to-noise ratio when interacting with most software is so low, I've just flat out stopped reading UI dialogs for the most part.

That's the pragmatic argument against excessive humanization, but there's another more visceral one that compelled me to write this in the first place: it's insulting to the very concept of personhood.

When Windows tells me "We're sorry, we can't do that right now", that's just a falsehood. Windows is not sorry, because Windows is not a sentient being capable of emotion. That message exists to trick me into thinking otherwise, because if the computer has feelings maybe I'll try to avoid hurting them by, say, not signing up for Microsoft 365 Active Copilot Pro For Workgroups (New) (& Knuckles). It's also just plain condescending.

Am I overreacting? Yes, probably, but nowadays it seems like every piece of software I interact with requires me to acknowledge its non-existent personhood before doing anything. Welcome, let's take a tour of our features! Deleting your account? We hate to see you leave! We found trending stories just for you! ². The LLM flood that's rendered large sections of the internet into useless slop has played a part in making this experience even worse, now that every website has an unhidable widget in the bottom right corner where you can ask a chatbot about the hours of this Red Robin and get possibly-correct answers in that chipper HR tone, but that's just the latest expression of the trend, not the origin.

I'm not religious, but something about this feels blasphemous. To be a person is to be in possession of an irreducible, unique perspective, unlike anything else in the universe. An unwavering band of light, fractally complex. Our language is an attempt to exteriorize that self. We cannot show others our actual experience of remorse, the thing-in-itself, but we can use words to point in its direction. Look, fellow unwavering band of light, this is what it's like for me, is it like that for you too?

When a SaaS widget says "We're sorry", the actual emotion of remorse is absent, of course, but it goes deeper than just a mere lie. We conscious beings can lie about internal states as well, but lying is still an action we take as conscious beings. If I say I'm sorry when I'm actually not, there is still some internal truth I'm referencing, even if that reference takes the form of contradiction. Those words coming from a modal dialog are in reference to nothing. It's not even a lie, it's just nonsense (or bullshit).

I know I'm not alone in feeling this way. I don't remember when I first saw this image, but the phrase "I am a divine being, you are an object" has become a personal mantra.

This is only going to get worse. Software acts like this because it's profitable for the corporations that fund their creation if we consumers think of our apps as our friends. I invite you, next time your email client says it's excited to show you its new AI features, to join me in recalling that you are a divine being; it is an object. It has no right to speak in your holy tongue.