The SensePath workflow supports an HCI researcher in recovering and understanding a participant's online (browser-based) sensemaking process from analytic provenance. It combines automatic provenance capture with interactive visual analytics to facilitate the transcription and coding phases of thematic analysis, turning a slow, manual video-transcription task into a faster, view-supported one.
The process begins during the participant's session, where the SensePath Chrome extension runs unobtrusively in the background and automatically captures sensemaking actions (analytic provenance) from browser interaction events. Search actions are detected by parsing URLs against per-service query templates; filtering is detected by a URL parameter-diff heuristic; reading, highlighting, and annotation are captured through a content script. Each action is recorded with four aspects: type, timing, context, and relationship.
The captured actions are arranged temporally into a timeline and presented through four linked views: a timeline view (overview of all actions as bars/tiles), a browser view (the web page of the selected action), a replay view (screen-capture video synchronised to the selected action), and a transcription view (a tabular, exportable transcript).
The analyst then enters an iterative transcription-and-coding loop: selecting actions in the timeline, inspecting their context across the browser, replay, and transcription views, interpreting the nature of each sensemaking activity (e.g., reading, comparing, locating), and assigning theme/code names. View settings (zoom, aggregation, length filter) are adjusted as needed to surface uncoded actions, and the loop repeats until coding is complete.
Finally, the analyst identifies recurring patterns across the coded timeline, characterises the user's overall strategy (corroborated by the replay video and think-aloud), and abstracts a model of the sensemaking process as an ordered diagram of named steps.