Pros and Cons of Controllability: An Empirical Study

Abstract

A key usability issue with systems that adapt to their users is controllability: the ability of the user to determine the nature and timing of the adaptation. This paper presents an empirical study of the trade-offs involved in an attempt to ensure a suitable degree of controllability. Within an adaptive hotlist for a conference web site, two mechanisms for providing users with recommendations of conference events were compared: automatic vs. controlled updating of recommendations. In an experimental setting, each of 18 users worked with both variants of the adaptive hotlist, as well as with a nonadaptive variant. The users differed markedly in their responses to automatic vs. controlled updating. A number of reasons for these differences could be found in the objective and subjective data yielded by the study. The study illustrates how preferences for different forms of user control can be influenced by factors ranging from stable individual differences to unpredictable features of a situation. General implications for the design of controllable adaptive systems are discussed.

The research described here is being supported by the German Ministry of Education and Research (BMB+F) under grant 01 IW 001 (project Miau). Björn Mittelsdorf and Marie Norlien made essential contributions to the empirical study.