Experimental evaluation of polite interaction tactics for pedagogical agents

摘要

The annual ACM International Conference on Intelligent UserInterfaces (IUI) is a unique meeting at the intersection of thefields of artificial intelligence and human-computer interaction.The ideas and methods from these two disciplines are appliedtogether to create innovations that promise to improve theusability and capabilities of our interactive computer tools. Thegoal is to enhance the performance of the human intellect throughpowerful computer interfaces that are a joy to use.>Every year the submitted papers span a wide range of topics,mostly concerning intelligent techniques for input interpretation,output generation, assistance, personalization, and interfacegeneration. Many papers describe the use and evaluation of thesetechniques in one or more specific application areas. During thereview process, papers are carefully vetted for the extent to whichthey demonstrate progress on questions that substantially concernboth artificial intelligence and user interfaces. Althoughreviewers may be sympathetic to papers of narrower scope, theyoften argue that such papers are better suited to a lessinterdisciplinary conference.>As intelligent user interfaces have proven themselves inpractical applications, the IUI conference series has seen adramatic increase in interest. This year, many more papers weresubmitted to the conference---and the reviewing process was moreselective---than in any previous year.>Fair and thorough reviewing is a challenging task for amultidisciplinary conference such as IUI, where no single personcan be an expert in the areas of all submitted papers. Thereviewing process this year comprised several stages: First, eachsubmission was read by three reviewers--in most cases members ofthe IUI 2005 program committee, but in some cases similarlyqualified special reviewers. Second, for each long papersubmission, a imeta-review/i was prepared by one ofthe especially experienced members of the program committee. Themeta-reviews were not just additional reviews: They analyzeddifferences among the evaluations of the reviewers, and in manycases they were based on discussion with the reviewers. Third, thereviews and the recommendations of the meta-reviewers werediscussed during a 14-hour on-line program committee meeting, aswell as in telephone conversations with meta-reviewers. Fourth, theremaining borderline cases were further analyzed by the program andshort paper chairs before they made the final selection of acceptedpapers.>Because of all of the effort that was invested in the reviewingprocess by so many people, we like to think that we have organizednot one but two IUI 2005 "conferences": In the first one, conductedvia electronic communication during the reviewing period, severalhundred authors of more than 200 submitted papers received carefuland often extensive feedback on their work from experienced peerreviewers. This feedback should help them to strengthen theirresearch and to prepare high-quality submissions to upcomingconferences and to journals. The second conference is the one heldin San Diego in January, 2005, whose proceedings you are nowreading.>For IUI 2005, 135 long papers were submitted, an increase of 40%over the previous record year for an IUI conference. In order tomaintain the IUI tradition of having each long paper presented in aplenary session, we had to limit the number of accepted papers tojust 30, for an acceptance rate of 22.2%. As a result, we had toleave out quite a few promising papers whose acceptance had beenrecommended by one or more reviewers. Each of the accepted longpaper submissions is printed as an 8-page paper in this proceedingsvolume. In addition, 7 long paper submissions that seemedespecially suitable for short paper presentation are presented inthe short paper track. For the short paper track itself, 74submissions were received (again, a 40% increase over the previousrecord year), of which 21 were accepted, for an acceptance rate of28.4%. Each short paper is printed on 3 pages in this volume and ispresented at the conference during an extended interactive posterand demonstration session.>A glance at the table of contents will reveal the extent towhich IUI 2005 is a truly international conference. A statisticaloverview of the countries from which the papers were submittedwould understate the diversity of the origins of the authors, manyof whom come from countries different from those in which they arecurrently working. We hope that in the future more authors fromLatin America and Africa will submit their best work to IUIconferences, so that the remaining gaps in representation can befilled.>Aside from the paper presentations, IUI 2005 includes a numberof events that help to ensure a varied and stimulating conferenceexperience. (See the table of contents for titles and names.)>Three half-day workshops and three half-day tutorials wereselected from those submitted on the basis of reviews by members ofthe conference committee. Between them, the workshops and tutorialscover a broad range of topics of current interest in the IUI field.We encourage readers to consult the summaries of these events thatare included in this volume and to explore the links that theycontain to extensive material that is available on the world-wideweb.>The program also includes invited talks by three distinguishedthought leaders who were selected after extensive discussion withinthe conference committee: Stuart Card of the Palo Alto ResearchCenter; Justine Cassell of Northwestern University; and Barry Smythof University College Dublin and ChangingWorlds.>Another highlight is a panel discussion in which researchers andindustry representatives meet to debate, together with theconference participants, about ways of dealing with the currentusability crisis in high-tech home products.