Vision & Cognition Laboratory |
Mind Tracking Our research explores how machines can read people's minds. Our approach to this problem centers on two basic questions. First, how can we represent people's thoughts and intentions as a computational model? Our work emphasizes the development of such models in a cognitive architecture -- a psychological theory and computational framework that incorporates the numerous parameters and constraints of cognition and perceptual-motor behavior. Second, how can we relate a person's observed actions to this computational model? Our work centers on the concept of mind tracking -- mapping a sequence of observed actions with the most probable sequence of actions predicted by a computational model. In essence, mind tracking uses the computational model to "think" along with the person being observed: at some current state, mind tracking considers the set of possible thought/behavior sequences that could arise from that state, and then determines which sequence best matches the person's observed behavior. When we combine cognitive architectures and mind tracking, we can infer the continual stream of a person's knowledge, skills, preferences and so on, which can in turn be used to add personalized intelligent support to computer interfaces, automobiles, computer tutors, and many other systems. Primary Reference Salvucci, D. D., Mandalia, H. M., Kuge, N., & Yamamura, T. (in press). Lane-change detection using a computational driver model. Human Factors. Related References Mandalia, H. M. (2004). Pattern recognition techniques to infer driver intentions. Masters Thesis (Tech. Rep. No. DU-CS-04-08), Department of Computer Science, Drexel University. Salvucci, D. D. (2004). Inferring driver intent: A case study in lane-change detection. In Proceedings of the Human Factors and Ergonomics Society 48th Annual Meeting (pp. 2228-2231). Santa Monica, CA: Human Factors and Ergonomics Society. Salvucci, D. D., & Gray, R. (2004). A two-point visual control model of steering. Perception, 33, 1233-1248. Salvucci, D.D., & Siedlecki, T. (2003). Toward a unified framework for tracking cognitive processes. In Proceedings of the 25th Annual Conference of the Cognitive Science Society (pp. 1023-1028). Mahwah, NJ: Lawrence Erlbaum Associates. |
We read people's minds countless times a day. We see a friend wave,
or hear a colleague ask "Coffee?", or see a person wink from across
the room, and we interpret these actions to infer the thoughts and
intentions behind them. Machines, like us, continually read people's
minds. Mind reading by machines is sometimes rather mundane -- for
instance, a toaster interprets a lever press as a command to toast.
But it can also be extremely subtle and complex: speech and handwriting
recognition systems interpret noisy signals as words; intelligent
vehicles interpret driver behavior to provide warnings and support;
e-commerce web sites track customer purchases to infer future buying
interests; tutoring systems interpret student responses to estimate
current knowledge. In these and many other cases, machines infer
the intentions behind observed actions in order to generate appropriate
responses. Mind reading is thus an essential component of communication
between humans and machines alike.