Discussion on Profile Structure Point13

From wiki.gpii
Jump to: navigation, search

In a user preference profile instance, each entry (property-value pair) may have a probability assigned

Status: Open under discussion (before the 3 April 2012 call, status was Approved approved)


This is a proposal for attaching probabilities to preferences in a user's profile. Probabilities would help software distinguish between preferences that have been input or confirmed by the user and preferences that have been inferred from other preferences. A probability is a value in the range [0; 1]; 1 means that the preference has been input or confirmed by the user; a value lower than 1 means that the preference has been inferred from other data. (Example of inferred preference: if a user wants a 16 pt font size on his PC, software could infer a matching font size for the user's smart phone. Until the user has confirmed that they really want this font size, the probability will be less than 1.)


Should probabilities be stored in the profile (as stored on e.g. a preferences server) or only used internally (i.e. in software that processes preferences)? In other words, should probability be internal to the "matchmaker" that makes the inferences? (Different matchmakers may assign different probabilities, e.g. because they use different mechanisms to make inferences.)

See Also