Continuing on the subject of context …..
[…] the major different cognitive structures of individual users […] may be represented to intermediary mechanisms and IR systems, e.g. the actual work task or interest leading to a current cognitive state which may end up in a problem or uncertainty state for the actual user. These mental activities take place in the context of a domain(s) of epistemic, social or organisational nature which, in addition to influencing the current searcher in a ‘historical’ socio-semantic sense […] also constantly influences the authors of texts and systems design attitudes. The most simple form of a domain is an academic subject field, which is essentially a social construct represented by the collective cognitive structures of the individuals forming the field. Other forms of domains are, for example, industrial sectors, individual firms or organisations, or professional groupings, such as journalists. The horizontal interactive processes […] display communicative properties that are different from the vertical ones because of human interference […]
Peter Ingwersen; Cognitive perspectives of information retrieval interaction: elements of a cognitive IR theory; Journal of documentation; 52:1; 1996; pp 3-50 (pdf)
Interestingly enough, Ingwersen links context to cognitive structures which, inter alia, consist of the social surroundings of the user, such as her organisational and professional framework; as well as her current work tasks and interests. He suggests that these structures should be presented to and utilised by the IR system in order to improve retrieval.
Context, in other words, is more than taking into account previous search behaviours, be it my own or others. It rather means building a workflow-dependent framework for the user and gathering meta-data about her current information task, as opposed to purely relying on previous search experiences.
[NB] See also my earlier posts on Keywords vs context, Context and association, and Law as an App.
iinek's blog 
If you believe the extended cognition literature* (and I certainly do), the IR system (or indeed any artefact) can be an integral part of the cognitive process. The design of the system is of course crucial in determining what kind of a role it plays in the process, whether it is a useful part or not. A hypothesis of mine is that this is particularly and increasingly important for the legal domain simply because of the volume, detail, and volatility of legal sources: most lawyers are more and more dependent on these systems in their daily work.
* Starting with Edwin Hutchins’s Cognition in the Wild (1995), and, just for equal measure for the opposing viewpoint, Adams & Aizawa’s The Bounds of Cognition (2008). For example Andy Clark has several books on the topic as well.
I could not agree more, Anna! Thank you also for the literature tips!