Amount of Information Presented in a Complex List: Effects on User Performance

Amount of Information Presented in a Complex List: Effects on User Performance Dawn Dutton AT&T Labs - Research 180 Park Avenue D103, Bldg. 103 Florham Park, NJ 07932 +1.973.236.6522 dldutton@att.com Selina Chu Information and Computer Science 444 Computer Sci. Bldg. UC Irvine Irvine, CA 92697 +1.949.509.9762 selina@ics.uci.edu James Hubbell 12 Lawrence Avenue West Long Branch, NJ 07932 +1.732.728.9351 jameshubbell@hotmail.com Marilyn Walker AT&T Labs - Research 180 Park Avenue E103, Bldg. 103 Florham Park, NJ 07932 +1.973.360.8956 walker@research.att.com Shrikanth Narayanan Speech and Image Proc. Institute Department of EE-Systems EEB 430, USC Los Angeles, CA 90089 +1.213.740.6432 shri@sipi.usc.edu ABSTRACT AT&T Communicator is a state-of-the-art speech-enabled telephony-based application that allows the end-user to, among other things, select and reserve airline itineraries. This experiment explores how the amount and structure of information presented in complex lists influences the user experience and the ability of subjects to successfully complete a selection task. Presenting all the relevant information needed for a decision at once was the factor that most positively influenced successful task completion and the user experience. Subjects preferred hearing all of the relevant information about each flight, without initiating additional dialog with the system. Additionally, successful task completion rates improved when all of the flights were presented at once, without any intervening questions from the system. Keywords dialogue, user interface, human factors, spoken language, lists, selection 1. INTRODUCTION AT&T Communicator is a state-of-the-art spoken dialogue system that allows the end-user to select and reserve various travel related resources, in particular, airfare, hotel, and rental cars [2]. One of its most challenging dialog modules is that for presenting information that allows the user to select a desired air travel itinerary. While selecting between multiple possible itineraries in a visual domain is a relatively simple task because most of the selection criteria are listed in a single scrollable page, the same task is likely to have a much higher cognitive load in an audio- only domain [1, 3]. The selection criteria for each candidate flight must be presented to the user serially, leading to higher cognitive demands that could result in errors in itinerary selection. A sample AT&T Communicator dialog with the prototype used in this experiment, showing the outbound flight only, follows. System: Hello, this is AT&T Communicator. How may I help with your travel plans? User: Yeah, I’d like to fly from Newark to San Francisco. System: OK, a flight from Newark, New Jersey to San Francisco, California. And will this be a round trip? User: Yes, and I need to arrive in San Francisco by 4 PM on November 22nd. System: I found five outbound United Airlines flights. Option 1 is flight number 635. It has one stop; it departs at 10:15 A.M. and arrives at 2:42 P.M. <pause> Option 2 is flight number 7535. It has one stop; It departs at 12:20 P.M., and arrives at 4:37 P.M. <pause> Option 3 is flight number 837. It' s a non-stop flight; it departs at 2:00 P.M. and arrives at 5:01 P.M. <pause> Would …

Content truncated to 500 words