Behavior analysis and prediction

Proposed by : Philippe Bossut

Skills : good UI understanding, statistic and algorithmics

This is a research project. It comes from the observation that users all do end up doing some similar things over and over again: emailing the same people, archiving or tagging emails coming from some mailing list a certain way or doing some peculiar tasks in a way different from the out of the box default (like creating only 30 minutes meetings instead of 1 hour). Traditionally, these specificities have been met by programs by supporting extensive preferences and providing users with tools to create scripts to automate repetitive tasks. Often non technical users simply ignore those tools and just end up building muscle memory... We're proposing here to develop heuristics that will detect these characteristics and adapt the behavior of the software automatically. This requires some level of instrumentation and logging to be added to Chandler, as well as statistical tools and heuristics to analyze those logs, detect patterns and predict user's behavior.