Today Eelco Wiechert presented and defended his master thesis on the occurrence of the peak-end effect in music playlist experience, supervised by Martijn Willemsen and myself. Psychological literature tells us that when evaluating an experience, all individual episodes that make up the experience have equal weights. When people evaluate an experience, they mainly consider the most intense episode (peak) and the last episode (end). Thus when we do our objective playlist evaluation, we should take this into account.
Eelco designed a very extensive study in the form of a Flask application running on Heroku building on Spotify‘s API that allowed participants to listen to personalized playlists that he manipulated to contain an extreme peak and end, or a smaller peak and end. The peaks and ends were done based on the valence and energy audio features provided by Spotify and in order to investigate the research question he asked users to rate all individual songs in the playlist and the playlist as a whole.
The basics of what he found was that the values of the most intense song and the last song indeed explain additional variance in the playlist evaluation over considering the average value of the individual songs. And there are (naturally) several smaller, interesting findings that can guide us in our future research.
If you’re interested in reading his work, I will add the link here as soon as it is available.