The document describes a study that uses play logs from online music services to infer a user's motivation for selecting songs. The researchers developed a topic model that considers both a user's usual taste in music and potential addiction to certain artists. Their model estimates the ratio of a user's selections driven by taste vs. addiction. Experimental results on two datasets show their model achieves lower perplexity than a previous model only considering taste. Qualitative analysis reveals insights like higher addiction ratios in mornings and on weekdays.