More and more personal information and artifacts of our individual pasts are available in digital form. A multitude of online services allow collecting and accessing information about one's life, replacing the formerly analog forms of such lifelogging. Despite the often-discussed privacy issues, millions of people spend their time recording their exercise and running habits, their moods and even their nocturnal activities. These services often enrich the available data with statistics and small graphs but their main use is to record and allow direct access.
Still, such lifelogging data can quickly become quite complex, especially when the information is no longer tracked manually by the user but automatically in the background. One such example is Last.fm, a popular webradio that promises to deliver personalized music by analyzing one's listening behavior. The most common way to do this is installing the Audioscrobbler, a demon process that follows the currently played music in media player software and sends this data to the Last.fm servers (scrobbles it). The by-products of this process, the recorded listening histories, are meticulous representations of one's music consumption and already have become the actual reason for many people to use Last.fm. They can, however, quickly span tens of thousands of songs and become too complex to be understood from the chronological lists that Last.fm provides.
A large number of fan-created static visualizations and analytic tools are therefore available that range from timelines displaying the number of logged songs (Scrobbling Timeline5), via high-level comparisons of multiple users to visually appealing Streamgraphs of the artists from a listening history. The main problem with these tools is that they might give an interesting and entertaining overview of the history but fall short in representing detailed information such as individual songs.
We believe that a 'casual information visualization' approach can prove valuable for making this personal information available to their creators. In this paper we analyze the data domain of listening histories and present our findings on their structure and what possible user tasks and available patterns it might contain. Additional contextual information can trigger the memory of the user to reveal the reasons for listening decisions. As a second contribution, we give an overview of LastHistory, a visualization of personal listening his- tories from Last.fm that not only allows sophisticated analysis of the underlying data in a non-threatening way, but is also able to show contextual information in the form of photos and calendar entries to help the user remember this time of his or her life. For this design study, we discuss the applied transformations for interface and views, user interactions and - with more detail - the differences between the two usage modes (analysis, reminiscing). We also evaluated LastHistory with four casual users and present anecdotal evidence for its value. A subsequent large-scale online evaluation with a corresponding questionnaire shone more light onto usability and feature issues. After making LastHistory available, several thousand people downloaded it and left generally positive feedback. We conclude with an explanation how the techniques used in LastHistory can be applied to other forms of lifelogging data and what other approaches might be fruitful.
Music Machinery: LastHistory : Visualizing Last.fm Listening Histories