
I recently volunteered to take part in a replicated experiment, originally performed in the paper ‘The unfriendly user: exploring social reactions to chatter-bots’ (A. Angeli, G. Johnson & L. Coventry; 2001.)
In that study, scientists were looking for how anthropomorphism would affect participants’ interactions with a chat-bot.
This title was presented to us before I was allowed to spend the week talking with ALICE and this genuinely affected my interactions.
I was certainly not going to be an unfriendly user.
Remember the AI twitter-bot from Microsoft? She got beaten within a day. I have never trolled anyone, and unless this was a particularly hostile bot, I wasn’t going to start. That being said, recent research suggests that under the right conditions — anyone can become a troll.
And I think they’re right.
Chat-bots were popular during college. I think it’s name was Oliver-bot, and I would spend hours with friends on separate machines putting difficult questions to it, checking if it could recall anything I told it and at times, being abusive just to see what it would say.
“While prior work suggests that trolling behavior is confined to a vocal and anti-social minority, ordinary people can, under the right circumstances, behave like trolls.” -Cristian Danescu-Niculescu-Mizil.
ALICE was programmed exactly like Oliver-bot, immediately transporting me back to my youth. The thing is, knowing my transcripts would be read I didn’t talk to ALICE in the same way I used to with Oliver-bot. I kept my emotions separate, didn’t insult or feel the urge to.
Not receiving a debriefing due to living far away from the session, I still wonder if the title of the original paper was given to participants to see if we would adhere to politeness during the chat sessions.
As far as anthropomorphism is concerned, my specific brain’s need to place a conscious value on androids and robots, combined with a love for Philip K Dick, Asimov and the like undoubtedly contributes to my interactions with anything that can communicate; living or otherwise.
But here’s the deal — we knew our transcripts would be read and analyzed, and would you really want people to know how much of a jerk you can be when nobody’s getting their feelings hurt?
I was certainly not going to be an unfriendly user — not with an audience. And maybe that’s as morally sound as humans will ever be.

