Agents and User Interfaces Marti Hearst SIMS 213, UI Design & Development April 29, 1999
Summary from Last Time People seem to treat interactive computers as if they were social actors What are the implications? for design of UIs for understanding social interaction?
Building Automated Agents Computer scientists trying to build believable agents: ignore the vast psychological literature on personality assume representations need to be rich need sophisticated natural language processing and intelligent interaction need realistic graphics, movement, and behavior
Anthropormorphia vs Ethopoeia Nass et al. distinguish: human-computer interaction is fundamentally social not anthropomorphic: “tending to believe computers are like people” human users behave as if computers were human, even though they know they are not ethopoeia: “assignment of human attitudes, attentions, or motives to non-human objects”
Laurel’s Definition (Brenda Laurel, Interface Agents: Metaphors with Character) Anthropomorphism in this context: Not the same thing as relating to other people Rather, the application of a metaphor metaphors draw incomplete parallels between unlike things emphasize some qualities, suppress others Two key anthropomorphic qualities wrt interfaces: Responsiveness Capacity to perform actions What aspects of human-human interaction are left out? Metaphor of Agency
Using Character to Depict Agency Drama and film capitalizes on our ability to draw behavioral inferences based on sparse character cues We can understand/enjoy even one-dimensional characters Stories are good only if out-of-character behavior can be explained causally
Using Character to Depict Agency Benefits of representing capabilities of agents using characters: leverages our abilities to make inferences about and predict likely behavior/choices invites conversational interaction doesn’t require detailed development of the agent characters in GUIDES interface have faces obscured; focus instead on period costumes, hair style, and surroundings can match the character to the user and/or the task
Agents vs. Direct Manipulation Debates CHI 97 and IUI 97 Personified in Maes and Shneiderman Can also be seen as AI vs. HCI Main Issue: How much should/can be in the user’s control, how much done “under the hood” by software? Outcome: They agree on a middle ground, that is more user-centric and user-driven.