Overview
The ever increasing quantities of information that people are required to deal with means the managing and re-finding information is becoming increasingly difficult. By studying how people use PIM tools we can understand what interactive support people need when re-finding, evaluate the effectiveness of existing tools and inform the design of more useful tools. This project sponsered by the Alexander von Humboldt foundation is aimed at developing new means to evaluate and learn about Personal Information Management (PIM) behaviour. The project is being run by Dr. David Elsweiler who is based at the Chair for Artificial Intelligence at the University of Erlangen-Nuremberg for the duration of the project. The aim of the work is to build mathematical models of how people interact with their email clients. An example of such a model is pictured below.
This is achieved by constructing a set of states, which represent user interactions, and counting the instances where the user moves from one state to another. In the model above the states are click-level interactions such as sorting, selecting messages, opening folders etc..
Generating this kind of model offers researchers several advantages. Firstly, what they provide is essentially a picture of the user's behaviour, which can be used to analyse and understand behaviour in ways not possible when observing participants live or by analysing log files using traditional techniques. Further, models can be generated based on the data for different kinds of users and in different situations and can be compared statisitically to determine if significant differences occur in these different scenarios. Another use for the models is to use them as a means to triangulate the findings of different kinds of studies by looking for examples of standard behaviours in naturalistic logs. The models could also be used as a means predict behaviour and offer appropriate support based on the user's current needs, forming the basis for intelligent email search tools. The models could also form the basis of simulated behaviour, allowing researchers to test search algorithms without the need to conduct full user trials. Developing these techniques will be extremely useful to researchers and potentially very beneficial to end users in the future. We are currently preparing a long-term study of email management and re-finding behaviour, which will gather the data we need to construnct such models. We are looking for participants to take part. If you would like to participate in the study click here.
Publications associated with the project
- Elsweiler, D.; Hacker, M. & Mandl, S. (2009), Visualising PIM Behaviour with Markov Chains., Personal Information Management Workshop at Annual Meeting of American Society for Information Science (ASIST 2009). Vancouver, Canada.
- Elsweiler, D.; Azzopardi, L. & Mandl, S. (2009), Evaluating Personal Information Management Behaviour using Markov Models., in proc. 5th Conference on Professional Knowledge Management', Personal Knowledge Management Track. Solothurn, Switzerland.