The traditional purpose of observing interactive behavior in usability engineering practice and usability research is to make inferences about the way users struggle with interaction elements to reach their goals. Several sources of data can be collected (e.g. video capturing, user interface events, voice, time measurements, navigational paths, eyetracking data) and have to be combined and interpreted in synergy to develop an interaction picture which serve the needs of the investigation. Beyond the need of identification of user interaction problems, the interpretation of data should lead to the prioritization of interaction problems (which has shown as a must in industrial projects), the identification of improvement potentials for the particular system, improvement and innovation potential in general and the usage of the data for advancing the knowledge of field (interpretation and comparisons towards the consolidation of "old" and "new" knowledge).

This leads to several interpretation goals and paths. In an ideal world every interpretation within an engineering or research project of interactive behavior data acts as an additional piece of a puzzle enhancing the existing understanding of interactive behavior. This leads to interpretation challenges on several levels:

• Data context: data is mainly collected on a very detailed level (e.g. event logging, eye tracking data) which does not efficiently support the interpretation in terms of interaction elements and interaction dynamics (e.g. co-ordination of eye tracking data with dynamic web content is expensive), data collection support based on higher level interaction characteristics (statics as well es dynamics)
• Time: sychronization of different interpretation sources is an extremely time consuming task which leads to evident problems in short termed (mainly industrial projects) user experience experiments where interpretation and prioritisation is the primary input to ongoing projects, furthermore working with designers iteratively needs continous rather than discrete interpretation of user experience data which also leads to a further level of interactive behavior understanding in terms of ongoing design changes and their effects
• Observation context: advanced environments like augmented reality or mobile interfaces will lead to another level of improvements to be made; there is a growing number of co-operative and multi-channel systems to be studied
• capturing: capturing the interpretation in a form which can be used for different purposes also covering the interpretation rational is an open problem which has to be tackled; this is also a prerequisite for comparing and consolidating interaction interpretations from different sources
• Data accurateness: there are drawbacks regarding the precision of different data sources (e.g. eye tracking data)
• Dimensions of user experience: much more observation dimensions have to be taken into account to develop a more complete picture of interaction behavior; examples are aspects of trust, enjoyment or presence
• Knowledge context: interpretation of interaction data has much more value in the context of pre-existing interpretations

Above mentioned challenges will be discussed based on experiences from different projects in different contexts.

© 2002 Noldus Information Technology bv