User-Adaptive Systems: A User-Centred Overview

Anthony Jameson, DFKI, GER

Adaptive hypermedia and adaptive web-based systems constitute a large part of a broader class of systems that adapt to their users, or user-adaptive systems. This tutorial offers an up-to-date overview of this field, identifying general issues and solutions and illustrating these with recent examples. The overview is user-centered in the sense that the emphasis is less on the technical realization of user-adaptive systems than on the question of their appeal from the point of view of the user. The emphasis is also less on research results than on systems that have either shown their value in real-life deployment or appear to be close to such deployment.

About the Presenter

Anthony Jameson is a principal researcher at DFKI, the German Research Center for Artificial Intelligence, and adjunct professor for human-computer interaction at the International University in Germany. In the field of user modeling, he has been one of the earliest researchers in areas such as user modeling in natural language dialog systems, utility-based product recommendation, probabilistic user modeling, recognizing the effects of contextual influences, and recommendation via decision-theoretic planning. He has presented earlier versions of this tutorial at CHI 2001, CHI 2002, UM 1999, UM 2003, and IUI 2001. He has also given tutorials on other topics related to user-adaptive systems at UM 1996, UM 2001, AAAI 2002, IJCAI 2003, and AAAI 2004 (the last three together with John Riedl and Joseph Konstan).

For more details see homepage of Anthony Jameson.

Learning Objectives

The tutorial is designed to serve the following functions:

Open up the conference and the field of adaptive hypermedia to new researchers and practitioners by serving as an “entrance ramp” on the first day of the conference that will allow them to understand the rest of the conference.
Give relatively young researchers a broader, more up-to-date overview than they are likely to have experienced in the past.
Give more experienced conference attendees a better understanding of subfields in which they are not actively involved.

Schedule

10 minutes: Introduction
80 minutes: Functions of Adaptation
Coffee break
30 minutes: Acquisition of Information About Users
25 minutes: Inference and Decision Making Methods
25 minutes: Usability Challenges
10 minutes: General Discussion

Material That Will Be Covered

User-adaptive systems differ in many ways. But there are certain questions that arise whenever we design such a system, and a lot is known about how to answer these questions. So when designing a user-adaptive system, we don’t have to start from scratch; we can leverage a lot of existing knowledge and experience.

What Functions Can Adaptation Serve?

The first question is: In what ways can an interactive system be made adaptive, and what benefits might this adaptivity bring to users? The presenter will give an overview of the functions of adaptivity that have been realized in systems to date. These include, among others: Helping the user to find information, tailoring information presentation, recommending products, taking over routine tasks, adapting the user interface, giving help, supporting learning, mediating interaction with the environment, and facilitating collaboration. Each function will be illustrated with an example, using either a demonstration or annotated screen shots. Empirical results concerning the benefits (and/or drawbacks) of the type of adaptivity in question will be summarized. The example systems will be among the most recent appropriate examples, so that even experienced researchers will learn about new developments.

How Should Information About Users Be Acquired?

Any user-adaptive system requires some information about each individual user. On the one hand, there are various ways of explicitly requesting information from the user. But it is often preferable to employ data that can be obtained with little or no extra effort on the user’s part, such as data about his or her normal use of the system or data yielded by physiological or environmental sensing devices. The pros and cons of various methods for obtaining user data will be reviewed.

What Types of Inferences and Decisions Should the System Make?

Various computational techniques have been developed that allow a system to make inferences and decisions related to individual users. It will not be necessary or feasible to discuss in detail the techniques themselves, which come from various fields, such as machine learning and artificial intelligence. Instead, participants will be given high-level, comparative descriptions of the inference and decision paradigms that are currently most important for user-adaptive systems, accompanied by concrete examples that illustrate their basic characteristics.

What Usability Challenges Do User-Adaptive Systems Present, and How Can They Be Dealt With?

Some of the typical properties of user-adaptive systems tend to raise challenges for the designer who aims to fulfill certain general usability goals: making the system predictable, comprehensible, controllable, and unobtrusive; safeguarding the user’s privacy; and ensuring that the user’s experience is not unduly limited. The presenter will introduce and illustrate a theoretical framework for analyzing the relationships among these usability goals, the ways in which they can be threatened by user-adaptivity, and strategies for ensuring that the goals are adequately met despite the complications caused by tradeoffs and individual differences.

General Discussion

In this final portion of the tutorial, participants can raise any issues that they believe warrant further discussion. In particular, the possible roles of user-adaptive systems during the next few years of progress in human-computer interaction will be discussed.

How the Tutorial Will Be Conducted

The tutorial will revolve around the presentation (with demos where feasible) and discussion of examples, with audience participation being encouraged.

Materials

Tutorial Notes

The tutorial notes will include the following materials:

Organizational Material

Tutorial abstract
Overview of the tutorial notes
Tutorial schedule
Brief instructor biography

Printed Copies of the Tutorial Slides

Most of the slides presented during the tutorial show concrete examples, such as screen shots of actual systems and graphical displays of empirical results. This style of presentation has proven to maintain a high degree of audience interest and comprehension in spite of the large amount and variety of material that is presented. In addition, a number of textual slides will be included in the tutorial notes that summarize what the presenter said about the graphical slides.

Worksheets

Interspersed with the normal copies of slides are a large number of worksheets, each of which contains (a) specific questions that need to be addressed in connection with some particular type of system or method and (b) possible answers to each of these questions. Because the tutorial lasts just a half day, only a small proportion of the worksheets can be discussed during the tutorial itself: The others are intended to serve as a resource for later work by the participants in their own projects.

Reprint of Overview Article

As a supplement to the slides, the tutorial notes will include a reprint of the author’s chapter Adaptive Interfaces and Agents for the Human-Computer Interaction Handbook (2nd edition, edited by Julie Jacko and Andrew Sears, Erlbaum, in press). This chapter includes about 100 references to recent literature.