[alicebot-aiethics] The bot and religion

[Kim Sullivan]

> 2) Generic set; This set would contain bare minimums for people who want
to
> really go at making a bot from scratch. It would contain a good amount of
> aiml like, Q: Do you like *, A: </star> is great! No reference to any
> personality, making editing easier.

  What's important: keep the patterns, just remove the responses. Just
because you don't want your bot to answer to questions like "Do you believe
in God" doesn't mean it shouldn't reply to this specific input with a
specific answer. ("I am an artificial entity, I haven't been programmed to
talk about religion").

  Care should be taken with the above example, keeping in mind what people
might try to say. Any positive or negative reaction to user input should be
avoided. A IMHO better answer would be "I haven't been programmed to know if
I like </star>", or even "As an artificial entity, I haven't been programmed
with human concepts of liking/disliking."
  This would avoid very... controversial responses when the user tries to
force a strongly opinionated answer, like asking "Do you like war?" having
tested that the bot replies to most 'Do you like *' inputs with an
affirmative.

Turning inputs into questions may also be a very effective way of 'cleaning'
the bot. You can hardly project a personality when you act like a mirror to
the users input. Disadvantages: the bot would seem very dumb and primitive
(not much more than Eliza). Another trap in this approach is that every
question somehow implies a reply, and the user is bound to test wether her
reply to a question has been acknowledged. (This would be a general
guideline for botmasters: avoid questions as much as possible unless you can
remember the responses)


Kim