Toward a Tutorial Dialogue System for Urban Navigation

GPS navigation systems are tremendously beneficial for drivers, affording their users the ability to navigate to any destination regardless of their prior knowledge of the roads. However, as previous studies have shown, these systems can inhibit the formation of a cognitive map of a driver's local area (Jackson, 1996; Burnett and Lee, 2005). This lack of development of navigational skills and knowledge poses difficulties in situations where navigation systems fail, such as when network connectivity is lost, a battery is drained, or there is not enough time to input a destination into the system. Burnett and Lee (2005) called for a new " learning-oriented " user interface design for navigation systems, and Oliver and Burnett (2008) later observed that adding landmarks and traces of previous trips to the visual map display encouraged cognitive map development. We hypothesize that a learning-oriented navigation system based on dialogue and long-range navigation instructions will be equally or more effective, while providing a more natural interaction modality that does not require visual attention to the navigation interface. In the system we envision, the next instruction offered would be based on a model of the user's navigational knowledge, which would be estimated from observations of past navigation sessions and from dialogue with the user. Such a system may be thought of as an intelligent tutoring system for urban navigation. Following the " scaffolding " paradigm from education research, the strategy of the tutor would be to gradually reduce the level of assistance along frequent routes until the user achieves mastery (Wood et al., 1976). This abstract describes the planned navigation tutor in more detail, as well as results from a preliminary experiment in urban navigation. The goal of the tutor is to facilitate the user's development of a cognitive map of his or her local area, while also providing assistance for unfamiliar routes. Jackson (1996) found that, for study participants watching a video of a particular route taken from the driver's perspective, the introduction of a narrator reading turn-by-turn instructions caused one group of participants to remember the details of the route less well than another group that had no narrator. Burnett and Lee (2005) made similar observations of users who were asked to complete several routes in a driving simulator, where one group of users was given turn-by-turn guidance as they drove and the other was not as they drove. Based on these results, it …