Memory Strength Prediction Accuracy in Learner-Adaptive Software for Application of the Spacing and Testing Effects

• In Event: Poster Session 11
  In Poster Session: Machines and Digital Agents

Abstract

Intelligent study scheduler (ISS) software programs prompt learners to test themselves on previously-learned items, adapting inter-study intervals to maximize the spacing effect’s benefit. Investigating the accuracy of ISSs’ scheduling predictions can inform both theoretical understanding and practical application of the spacing effect. Using 265 Mnemosyne ISS users’ data, I estimated hierarchical ordinal models of memory scores across over 1.3 million item rehearsals. Analyses show that the Mnemosyne algorithm is moderately biased toward scheduling items too infrequently and does not account for large amounts of inter-item and inter-learner variance. However, the form of predictions is sound across varying interval lengths, consistent with an expanding retrieval model of the spacing effect. As expected, users’ adherence to the algorithm’s schedule predicted higher scores.

Author

• Matthew Lineberry, University of Illinois at Chicago