Conversational assistant for in-car task management

Abstract

The number of devices (in-car navigation systems, cell phones, email, web services, etc.) competing for a driver's attention is increasing creating the need for controlling cognitive load. We present a system, the Persistent Conversational Assistant, which strives to reduce cognitive load by providing the user with an intelligent-seeming assistant. Intelligence on the assistant's part can reduce the need for the driver to adapt to the system interactions. There are significant psychological factors to be addressed in building such a system. Since the in-car domain generally involves only a few users interacting with the system over a large period of time, the "bank teller" or "phone menu" style of interaction, though easier to design and implement, may actually increase user frustration, reducing the user's cognitive capacity. We posit that it is better to provide a "persistent assistant" that learns the drivers preferences and proactively simplifies common tasks. Copyright © 2004, American Association for Artificial Intelligence (www.aaai.org). All rights reserved.