In an earlier post, I described Waterworth and Chignell’s model of information exploration, and distinguished in from other theories ad models of information seeking in that it tried to address some aspects of interaction.The main problem with the model is related to the structural responsibility dimension. Structural responsibility models “who [system or user] is concerned with the structure [of the data]”, but since the user can only interact with the constructs exposed by the system through the interface, and the same structures can be expressed in many different ways, this dimension fails to capture a distinction that’s useful for design.
Interaction method is much closer to informing useful design decisions, but designers must be careful to use the dimension to design the interfaces rather than the underlying structures. For example, aspects that allow users to browse data on shopping sites are examples of referential interactions that use queries to select the data.
I propose a new model intended specifically to inform interaction and systems design for supporting information seeking. The goal of this model is to pick up where other models (e.g., those mentioned in the first post) leave off. The model consists of three dimensions: input method, atomicity, and explicitness.
Input method characterizes interactions as selection, manipulation, and production. Selection refers to such actions as link following, menu selection, list-box selection, or other interactions that constrain the user to select among a small set of choices. Manipulation extends selection by introducing the notion of how something was selected. Drag and drop interfaces, sliders, and other forms of near-continuous interaction fall into this category. Finally, production refers to input generated by the user, including query terms, freeform digital ink, spoken commands, etc.
The atomicity dimension refers to the extent that the output of the system in response to a particular command relies on earlier commands. For example, running a google query is an atomic operation, because each query is independent of the queries that came before. On the other hand, adaptive systems will build user models based on prior interactions with the system for the purposes of adjusting the results based on what the system thinks the user already knows or doesn’t know.
Explicitness reflects the user’s intent: did the user set out to seek information (explicit), or did the system react to user actions in some other context (e.g., reading or writing a document) by identifying potentially useful information (e.g, related documents)?
I’ll work out some examples in the next post in this series, and welcome comments and system examples to classify.