Towards a research agenda for long term HCI

Richard Thomas

ABSTRACT Most user studies have focused on either novices or experts, with relatively few reports of the progression between them. In this talk one of the longest investigations of user learning will be briefly reviewed. Over 4,000 individuals, mainly computer science undergraduates, were tracked in a seven year period. The main findings were that, for this group of users and software at least, activity seems to be constrained by three processes. These are: continuous low-level exploration, rather stationary frequency distributions of command invocation and consequent effect on the Type-Token Ratio, and sudden crossovers between methods.

Three new lines of enquiry that have become of interest as a result of this research and will be suggested. First is the opportunity for better statistical modelling of exploration profiles based on observed Type-Token Ratios and distributions. Techniques from information retrieval might be applicable here. Second is the identification of meta skills for long term learning and whether they should be taught to users. Finally there is the possibility that a technological underclass may develop because of the very long learning times in HCI, and whether it matters.

Richard Thomas
On leave from Department of Computer Science The University of Western Australia richard@cs.uwa.edu.au http://www.cs.uwa.edu.au/~richard