New Paradigms for Adaptive Interaction
Half-Day Course offered at CHI 2009
Boston, Thursday morning, April 9th
Krzysztof Gajos and Anthony Jameson

Introduction

With the term adaptive interaction, we refer to interaction with systems that automatically change their appearance, functions, or behavior on the basis of information about an individual user or group of users; this information may range from long-term properties of the user such as abilities, impairments, and general preferences to situationally variable factors such as short-term interests and other activities that the user is currently engaging in.

Learning Objectives

1. By taking part in the course, participants will acquire a well-structured understanding of a number of representative developments in the area of adaptive interaction which (a) go beyond systems and methods that they are likely already to be familiar with and (b) provide useful examples that can be followed by those who are interested in designing, evaluating, or deploying systems with adaptive interaction in the near future.

2. Because of the systematic way in which the material is presented and discussed, participants will also gain an active understanding of a set of concepts and principles for thinking about adaptive interaction.

Justification

New application scenarios for adaptive interaction, coupled with advances in research in this area, have led to a number of exciting developments that deserve the attention of members of the CHI community. And indeed, there has recently been an increase of interest in adaptive interaction within this community. For example, at CHI 2008, an unprecedented number of papers in this area were presented, including one of the 7 that won the CHI 2008 Best Paper prizes.

On the other hand, it is not easy for members of the CHI community to acquire up-to-date information on adaptive interaction that meets the requirements of being both new and practically applicable. On the one hand, the most widely deployed adaptive systems (such as the Smart Menus of Microsoft Office or the web pages of amazon.com) are so familiar that many CHI attendees prefer to hear about newer approaches. On the other hand, much of the research literature found in specialized journals and conference proceedings describes developments whose practical deployment lies at some unknown time in the future. Some of the information that would be of greatest interest to CHI participants is found in inconvenient sources like patent applications, white papers of companies that develop and sell adaptive systems, and weblogs and personal communications of researchers and practitioners.

The complementary and extensive experience of the two instructors of the proposed course puts them in an unusually good position to access the sort of information that CHI participants find most useful and to present it in a way that is meaningful for such participants.

Basic Structure

The course will be structured around the presentation of six currently significant paradigms of work on adaptive interaction. By paradigm we mean a method for adapting interaction in a particular way for a particular purpose that can serve as an example for researchers and practitioners. Each of the paradigms has been realized, tested, and in some cases practically deployed in one or more systems. In each case, we will use one or more of the relevant systems to illustrate the paradigm and enable concrete discussion of it. A set of paradigms has been chosen that is representative of the variety of forms and functions of adaptation that are currently being worked on. Each paradigm represents significant progress relative to previous work that participants are likely to be familiar with; and in each case there is evidence available concerning the added value of adaptation relative to nonadaptive methods.

Questions to Be Discussed With Reference to Each Paradigm

Each paradigm will be discussed according to a basic schema that is intended to help participants think systematically and realistically about adaptive interaction.

1. How does the paradigm look in practice?

Wherever possible, we will offer a live system demonstration with audience participation – or at least some other vivid and realistic presentation of example systems – so as to engage participants in the critical discussion of the paradigm in question.

2. What empirical evidence and/or practical experience is available that tells us about the strengths and limitations of this paradigm?

Evaluations of adaptive interaction sometimes involve methods that are not commonly used with other types of user interface (e.g., methods for evaluating the accuracy of a method for predicting a user’s behavior). Where a study about one of our example systems has some such interesting methodological feature, we will call attention to it and characterize it in a general way, so that participants can make use of the example when planning evaluation studies of their own.

3. What is the added value of adaptivity in this paradigm, relative to nonadaptive approaches to the same problem?

When it comes to the practical deployment of adaptive interaction, a crucial question is often that of whether the benefits yielded by the system’s adaptivity, relative to what could be achieved with nonadaptive methods, are great enough to justify the additional costs of adaptivity; the possible costs can range from usability problems associated with the adaptivity to costs of implementing and maintaining an adaptive system.

4. What are the advances over previous related forms of adaptive interaction?

To put the discussion of each paradigm in a larger context and improve the participants’ understanding of the field of adaptive interaction as a whole, we will briefly indicate the paradigm’s relationship to previous adaptive approaches to the same general problem (though there will not be enough time to discuss earlier approaches in detail).

Background of Attendees

The course is aimed mainly at participants who have at least some familiarity with the basic concepts of adaptive interaction and with a few examples of such systems. This condition should be fulfilled by most CHI attendees who are interested in enough in this topic to consider taking the course in the first place. For attendees who have exceptionally little knowledge of adaptive interaction, the course could serve as an introduction to the field, though these participants would probably not fully understand some of the more advanced discussion in the course. Even those who are experts on some particular type of adaptive interaction may choose to attend in order to learn about recent developments on other parts of the field.

This course should not be attended by those who desire a treatment of the technology underlying adaptive interaction, since this topic would require a separate course of its own. The technology in the systems discussed will be sketched only on a very high level.

Presentation Format

The most important information will be conveyed by the instructors by means of system demonstrations and brief lecture segments. Although the content of the course does not offer good opportunities for separate hands-on work sessions, the participants will continually be encouraged to express their judgments on the various points being discussed and to report briefly on relevant experience of their own.

Course History and Presenters’ Experience

Although the two instructors together have more than three decades of experience in research, industrial practice, and instruction in the area of adaptive interaction, they have now collaborated for the first time to design this new course for CHI 2009. The first instructor has made numerous presentations of his own research in this area at CHI and other leading international conferences. The second instructor likewise has many years of experience with conference presentations on the same level, as well as with relevant industrial and university courses and with tutorials at CHI, IJCAI, AAAI, and the more specialized conferences devoted to adaptive interaction.