I've written a number of "flash-card" style vocabulary learning programs over the years that have attempted to "adapt" based on the user's performance.
The algorithms for determining which card to show next have been largely ad hoc, but I'd now like to give them a more statistical basis. If we assume that the card we wish to show next is the one for which the time since last asked divided by the estimated time to forget is greatest, how should the estimated time to forget be calculated? I am trying to get up to speed on the relevant statistical theory to apply here. It seems that survival data statistics might be a promising approach but I'm not sure how exactly to apply it to individual flash cards. Can anyone help me? To make it more concrete: imagine that I have n flash cards and for the jth showing of the ith flash card I record t_i[j], the time asked, r_i[j], the result (right or wrong). Thanks in advance James . . ================================================================= Instructions for joining and leaving this list, remarks about the problem of INAPPROPRIATE MESSAGES, and archives are available at: . http://jse.stat.ncsu.edu/ . =================================================================
