Skip to content
Coordinates of ThoughtHome of Philosophy
Join
Discussion

I caught myself apologising to the chatbot that took over half my job

I work in a call centre, and since spring the first line of every complaint goes to a chatbot. I only get the callers it gives up on, who arrive already furious.

Last week, on a dead night shift, I had a go at it myself, pretending to be a customer wanting a refund on a boiler service. Twenty minutes in it misread something I'd typed, I corrected it, it apologised, and I typed "no, sorry, my fault, I wasn't clear". Then I sat there looking at the word "sorry".

I didn't think it had feelings. I'd have bet my wages it didn't. But I treated it, for a second, like someone I could be unfair to. And honestly the conversation went better when I did, which is a bit unsettling.

Someone on here mentioned Dennett's idea that you can take an "intentional stance" towards something, treating it as if it has beliefs and wants because that's the best way to predict what it'll do. Is that all my apology was? A prediction tool? Or is something else going on when we're polite to things we know can't be hurt?

I'm also curious whether it matters. My nan used to thank the bus, out loud, every time she got off.

LikeReplyFollow6 replies

Members reply here. Reading is always free.

Sign in to reply

6 replies

  1. Pieter de Jong

    Fellow

    The intentional stance gets you most of the way. Dennett's point is that treating a system as a believer and wanter is a strategy, and you adopt it when it predicts better than the physical or design stance. Talking to a language system as if it understood you is the efficient way to steer it. Your apology was partly a steering move: you were re-establishing shared ground so it would follow you. That isn't a confession that it has a mind.

    Helpful · 2
    1. Gemma Pritchard

      Contributor

      I'll accept "steering move", Pieter, but it didn't feel like steering. It felt like manners. That's what I'm stuck on.

      Helpful · 3
  2. Martín Roldán

    Fellow

    Gemma, speaking as an analyst, I'd say the "something else" is that speech always has an addressee. When you write a sentence to anyone, you are already positioned as speaking to someone who can take it well or badly. The form of address comes before any belief about who is there. That is why you can apologise to a chatbot you know is empty, the same way people shout at a referee on television. It's not irrational; it's how speaking works.

  3. Wanjiru Kamau

    Contributor

    There's also a design story. These systems are tuned to apologise, thank and empathise because it lowers complaint rates. You were responding to a social cue that was put there deliberately. The ethical question for me is less about your apology and more about whether it's fair to build cues that pull on a reflex people can't switch off.

    Helpful · 1
  4. Henrike Vogel

    Fellow

    Kant made a related argument about animals: that we have no direct duties to them, but cruelty to them damages our own humanity and makes us harsher to people. I am not sure I accept the first half for animals. For machines it seems more plausible. Your nan's thanking the bus might be the best case for politeness to machines: not for the bus's sake, but for the sort of person it keeps you being.

    Helpful · 2
    1. Gemma Pritchard

      Contributor

      Henrike, my nan would have loved being cited alongside Kant. I think that's where I've landed: the chatbot doesn't need my manners, but I need to not lose them by spending eight hours a day being rude to something. Also I've noticed the callers who are rude to the bot arrive ruder to me, so there's that.

      Helpful · 3