I am interested in the question of how people use/integrate previous experiences with instances of tasks or events to make predictions about the duration of future instances of tasks/events.
To illustrate: I am a regular user of Google Calendar, and I use its drag-and-stretch time-interval interface. In doing so, I constantly seem to be making judgments such as "Event A has properties [a1,...,aN], ####.... it will probably last about two hours". My question, then, is: what is happening during the "####" stage of this process?
**EDIT: To avoid confusion, I would like to make clear that I am interested in cases where an agent is facing an "instance" of a "type" of event (if you imagine that the agent has a set of "memory traces" related to instances of the event, then I am referring to a case in which there is a reasonably high amount of variance/uncertainty in duration across previous instances), that agent has certain knowledge/evidence, and the agent must make a guess and/or a decision based upon a guess. I've used the calendar planning example as an illustration. However, I am asking this question more broadly, as it relates to time scales ranging from very small to very large.
This question can be addressed at many levels of analysis (in Marr's sense). I am familiar with research related to this question, such as Griffiths and Tenenbaum's "Optimal Predictions in Everyday Cognition", Warren Meck's "Neuropharmacology of Timing and Time Perception", and empirical work on human planning behaviors/suboptimalities.
Can you recommend any further readings on this topic? I am particularly interested in research that treats this problem as an inductive reasoning task and investigates the question through the use of computational models.
UPDATE: I have just begun finding some very relevant articles, such as these:
1) http://www.ncbi.nlm.nih.gov/pubmed/15716368 (link to abstract)