I attended an interesting lecture this morning at UBC by Dr. Barbara Grosz, Dean of the Radcliffe Institute for Advanced Study and Higgins Professor of Natural Sciences, SEAS Harvard. Dr. Grosz is an expert in human-computer interaction (HCI) and gave an abbreviated version of tomorrow’s Faculty Associates Dinner talk at the Peter Wall Institute entitled, “Can’t You See I’m Busy? Designing Computers That Interrupt Only When They Should.”
I met my classmate Sara outside the lecture hall and we entered with a few minutes to spare only to find, with some dismay, that there was exactly one other audience member. What kind of talk would it be if we were an audience of three? That’s not fun.
With some trepidation I fetched a bottle of water from the refreshment station and we sat down and spent a few minutes planning our summer 2010 Project Euler challenge. I’d forgotten about the just-in-time mentality which seems to be in favour on campuses and particularly in comp sci faculties these days. With mere seconds to spare, I looked up and realized the room had filled. I’m not kidding. It was organized, unchaotic, quiet and polite, but in two or three minutes we went from an audience of 3 to hundreds. That was more like it.
After a brief intro Dr. Grosz launched her motivation for research with some very funny examples of computers interrupting the user with useless or redundant information. Message windows that pop open and say, “Mouse not working. Please check your connection and left click to continue”, or who can forget that useless Office Assistant from MS Office, Clippy the Useless ClipArt PaperClip? “I see you’re digging a hole. Is that a personal hole, or a business hole?”.
Clippy the MS Office Assistant (97-03)
There was a lot of groaning when Dr. Grosz brought up Clippy, though Sara who is about 15 years younger than me looked charmingly confused. But the question was clear: how can we design an artificial intelligence that will collaborate and not just interact? How do we design a computer that interrupts us in ways that we will find meaningful and helpful and not just distracting or irritating? How do we measure and map the utility of interrupts?
Dr. Grosz encouraged the room to review papers by Doug Engelbart, the inventor of the mouse, specifically the ‘classics’ (my quotes) from the 60s, in order to provide a conceptual framework for the problem domain.
Dr. Grosz introduced us to the Colored Trails Testbed, a framework for conducting research about decision-making in groups comprising people, computers and a mix of the two. It’s a simple game: a rectangular board of colored squares in which each player is given a starting position, a goal position on the board, and a set of chips in colors taken from the same palette as the squares. In order to attain the goal, the players have to bargain with each other.
Colored Trails
I was particularly interested in the overlap with anthropology and sociology that her work takes. Dr. Grosz commented on the difference that a financial scoring matrix played, and the differences between person to person and person to computer interactions and bargains. She examined notions of trust, benefit and cost, altruism and uncertainty, and of course collaboration.
Unfortunately the 90-minute talk was shortened to less than an hour followed by a brief and tightly-scheduled Q&A period. Dr. Grosz skipped through many of her slides though I caught tantalizing glimpses of graphs plotting experimental statistics and HCI buzzwords. Her concluding remarks boiled down to her observation that we need computer agents that are partners rather than servants, who will use context and behaviour to help us in meaningful ways. I am interested in reading more about her SharedPlans model of collaboration developed in collaboration with Sarit Kraus.
