Recognizing User Behavior from Interactions for Adaptive Consumer Information Systems

Abstract

Consumer Information Systems, which experience widespread application, benefit substantially from adapting the conveyed information to specific user needs, by addressing various impairments such as color blindness, deficient preknowledge, and/or graph illiteracy. Ideally, to allow for an unperturbed exploration process, the system automatically recognizes and responds to the need for adaptation. While it has been shown that users' interactions with a system can be leveraged to this end, there exists no generalized taxonomy covering all possible interactions/processes and how they relate to each other. This paper garners different interactions, defined in the literature, and classifies them regarding complexity and inter-dependencies in a 'processes landscape'. Using this landscape, we outline a concept how low-level interactions (e.g., 'Clicking', 'Typing') can be combined with context-sensitive ones (e.g., 'Hovering') to estimate high-level behavior such as 'Reading' or 'Exploring'. Knowledge of the latter allows a system to intervene and adapt in a reasonably manner.