Person or robot?

Geoff Pullum on the Lingua Franca blog today, in a posting entitled “Away from one’s desk”, about vacation messages:

Ages ago, when e-mail was young and did not yet dominate our lives, and vacation response was not a built-in feature of mailers, and the Berkeley Unix /usr/bin/vacation had not been written, a friend of mine at Stanford (the linguist Arnold Zwicky) had a little vacation-response script written for him in the Bourne shell language by a computer staff person (it was remarkable how little code it took), and he chose a first-person message. Immediately people started replying to the script: “I know you say you’re away, Arnold, but if you could just take a minute to look at this … “: They could not grasp the idea that a shell script had mailed them. The innate human tendency to perceive agency was too strong. If it said “I am away,” they thought that was Arnold talking.

So I always use the third person.

People are very much inclined to treat computer programs and their products as the work of sentient beings, a fact demonstrated many years ago in the way people interacted with the ELIZA program.

In this case, I continued to use the first person (because I was uncomfortable splitting myself into two entities), but added a note saying that the message was coming to you from a program, not a person. and that it was pointless to reply to it. That seems to have worked.

On ELIZA, from Wikipedia:

ELIZA is a computer program and an early example of primitive natural language processing. ELIZA operated by processing users’ responses to scripts, the most famous of which was DOCTOR, a simulation of a Rogerian psychotherapist. Using almost no information about human thought or emotion, DOCTOR sometimes provided a startlingly human-like interaction. ELIZA was written at MIT by Joseph Weizenbaum between 1964 and 1966.

ELIZA was incredibly primitive (see details on the Wikipedia page), but people engaged with it intensely.

As a bonus, Geoff supplied an entertaining xkcd cartoon on university websites:

Venn diagrams are all the rage.


5 Responses to “Person or robot?”

  1. arnold zwicky Says:

    Brent Woo on Google+:

    Voicemail poses a similar question for discourse analysts, people cope with the lack of conversation partner by filling silences, no need to negotiate pre-closings, etc.

  2. the ridger Says:

    Have you seen the Euro Venn (from Paul Krugman)? I love it.

  3. the ridger Says:

    I tend to get more annoyed, by the way, with a phone-bot that gives itself a name and says “I” than I do with with a less personalized one that is also not understanding what I say. “I didn’t quite get that, would you say again your 16-digit account number” is infuriating. (Especially since the eventual actual human has NEVER had the account info passed to them, but that’s a different rant.)

  4. arnold zwicky Says:

    From Steve Wechsler in e-mail (posted here with his permission):

    Liz Coppock once suggested to me that ‘the innate human tendency to perceive agency’ when 1st person is used, and the resulting annoyance at machines using it, could be seen as evidence for my (Language 2010) theory of 1st/2nd person, according to which ‘I’ is not specified for reference *to* speaker but only for self-reference *by* speaker, so that the addressee can only recover that reference through ‘Gricean’ inference, which requires imputing agency. At least I think that’s what Liz said. If not then I’m saying it.

    The Language paper:

    Wechsler, Stephen 2010. ‘What “You” and “I” Mean to Each Other: Person Marking, Self-Ascription, and Theory of Mind.’ Language 86.2.332-365.

    Slides from a recent talk Steve gave with Eric McCready are here.

Leave a Reply

%d bloggers like this: