Selecting an Effective Bot Character
Published by Christy Dena February 5th, 2005 in Uncategorized, Poetics.What sort of character suits a chatbot? An inhibited and inarticulate character obviously does not. Characters don’t have to be eloquant though they can be abusive, like one of my favourites: Artificial Iniaes. What do they NEED to have as traits? Obviously, they need to want to talk, talk with strangers, want to ask, want to tell, be able to handle abuse, be interesting…What sort of person does this make? Janet H. Murray discussed this in Hamlet on the Holodeck all those years ago:
?One of the first things students learned in creating their own chatterbots was how to select a good character. The most successful characters have been those who are self-absorbed, evasive, or obsessive in familiar ways.? (Murray, 1997, p.219)
?In creating Salesman, the author invented a method of making the rules for a conversation serve as a dramatic structure for an interactive scene. This modest exercise is an example of how new narrative techniques can develop when a tradition of composition is supported by an authoring environment that does not require programming. ? Computer characters who can carry on persuasive conversations could be an expressive narrative genre in themselves, as well as compelling elements in a larger fictional world.? (Murray, 1997, p.222)
Murray, J.H. (1997) Hamlet on the Holodeck: the Future of Narrative in Cyberspace, Free Press, New York.
Dirk Scheuring, an (Alice) botmaster sees the relationship between character and effective design likewise as choosing a character that limits or defines the realm of possibility and provides the user with a goal:
‘The author?s perspective on Conversational Agents is based on the premise that they represent, not “chat machines,” but a novel Storytelling medium. “Stories,” in this case, are defined as goal-oriented Human-Computer Interactions; e.g. online sales transactions, “troubleshooting” conversations, etc. Stories of this type can be modeled using the Grand Argument Story model developed by Phillips/Huntley.[6] An Argument Agent, as represented by this Blueprint, uses a “table-based” strategy in order to “argue” with the User about progress towards the Story Goal, thereby dramatically (in two senses of the word) restricting dialog possibilities. This leads to the AA appearing much more “reasonable” to the User than the average Conversational Agent using a “resolution-based” dialog strategy.’
Scheuring, D. (date unknown) ‘Argument Agents: An alternative approach to writing conversational agents‘, Association of Knowledgework, [Online] Available at: http://www.kwork.org/White_Papers/agent.html
Are we creating a breed of obnoxious agents? No wonder they end up taking over the world! (Oops, wasn’t supposed to reveal that.)
Anyway, thoughts on character design for bots?
I don’t mean to sidestep here, but that is one of my favorite chatbot response tactics.
To combine some of these comments you quote, it seems to me that more important than the character of the bot is the context suggested by the character.
As in:
and here:
.
When I think of Saucy Jacky or Julia, I think about the cyberspaces in which one encounters them, the context of the conversation. The chatbot, written into its identity, must have, like an actor, motivation, except it must have motivation for the user. Why are we chatting with it? How is the user being assimilated into the encounter through the address of the chatbot?
In short to answer:
They need to make us want to talk to them.
Oh I agree with you that it is the narrative framing that is part of the immersive experience: who are they and why are they and why we’re there. We seem to need to have a reason for contact in real life and the motivation follows on to chatting with bots. But it is extreme motivation. To spend time chatting with an artificial character, with ergodic literature, requires a great deal of effort (perhaps only in relative terms because we’re used to entertainment being delivered and pre-configured). A good reason to talk powers our keyboard tapping for a few minutes, right? But just like a good story the characters and the writing need to come into play to keep the reader there. Characters are perhaps sustaining devices rather than hooks?
Also, the need to make us want to talk to them obviously starts before the conversation. Bots are pitched as characters at present — you go and talk with a personality (just about all bots have names and this is probably due to the community being Loebner Prize oriented rather than creatively). Many bots aren’t really pitched as a personality in a dramatic situation (like character comedy, a fish out of water type tensions). Good characters imply a story, like Saucy Jacky as you mentioned, and John Lennon (both from the same company: Persona Bots). There are some bots that are framed around an event though, rather than a personality. See Timothy Marsh’s Elevator Bot and another that I don’t have permission to cite as yet but which is an audition scene. This later addresses nicely your second point:
It has the user represented on the screen as a co-bot-actor. The users role is implied, therefore, visually and in the content with direct address from the bot informing the user what they are supposed to do. Elevator bot outlines the situation and you know pretty quickly what you’re trying to do. They’re both goal-oriented. These are, of course, the same issues that game designers face with integrating and representing the player. Some papers on first and third person integration in games would be good here…
Here’s one such reference:
Khoo, A., Zubek, R. 2002. “Applying Inexpensive AI Techniques to Computer Games.” IEEE Intelligent Systems. 17(4): 48-53.
Also, after reading a transcript between ALICE and Alan, I’m starting to be more convinced of the importance of a sense of humor, or at least a display of attitude. Of course, getting back to motivation, it helps that Alan has something he wants to prove.