IUI Conference Review
Intelligent User Interfaces Conference
San Diego, January 2005
by Jamie Kravitz
I learned a new phrase in San Diego: “embodied conversational agents”, or “ECAs.” According to the MIT press, “embodied conversational agents are computer-generated, cartoon-like characters that demonstrate many of the same properties as humans in face-to-face conversation, including the ability to produce and respond to verbal and nonverbal communication.”
Many of the lectures and demonstrations at the conference featured these artificially intelligent devices that interacted with humans in a multimodal fashion. These lectures included a talk by Justine Cassell, creator of REA, an artificial real-estate agent who is apparently the biggest star in the world of ECAs. There were less sexy ones, too, like a fancy-pants thermostat, and the “intelligent” bathroom mirror that let you shave and program the lights over the toilet at the same time (I kid you not!).
To tell the truth, I was in a little over my head. The focus of the conference was pretty heavily into artificial intelligence (AI), and many of the attendees were hardcore computer/AI types. However, there were several panels specifically focusing on usability and other issues that relate to interaction design.
It was good exposure to the whole issue of multimodal input, as many of the programs, objects and gadgets on display used natural language, gestural input, and various types of tactile input.
Some of the presentations and posters I found particularly interesting: using geometric pattern matching to recognize stylus input with an on-screen keyboard; embedding hidden messages in background music that could only be perceived by people who were sensitized to hear it; a program for generating walking directions in a confusing building; and the conference-honored “outstanding paper,” which used an intelligent algorithm for media allocation, determining whether text, graphics, or voice (or combo) would be most suitable for the request the user made.
On the usability front, Barry Smyth from University College Dublin talked about privacy and personalization. He noted that while personalization can increase the usability of products, there is a trade-off with regard to privacy. There was also a panel discussion entitled “The Usability Crisis in High-Tech Home Products,” which was intended to spur discussion of how intelligent user interfaces could help products with poor usability.
Interestingly, it was not clear to me that some of the “intelligent” solutions offered made the products any more usable. In some cases, like the intelligent bathroom mirror, they seemed to add a lot of functionality, but were no more user-friendly than the poorly designed products they were supposed to replace.
The conference was small, with only one track, so there was ample opportunity to meet with others in the field and connect with the presenters. While a little esoteric, there was plenty of interesting information for HCI-oriented human factors people, and interesting glimpses of cutting-edge research in the artificial intelligence community.
