The LifeSampler: Enabling Conversational Documentary

Summary

I use the term "Conversational Documentary" to refer to forms of documentation that encourage many-to-many relationships among content creators and viewers.  These technologies empower viewers to become creators in turn, generating new content that is positioned in a web of meaning with existing content, adding value for both prior contributors and future viewers.  I emphasize video in my own work, partially due to my background in film and partially due to the familiarity of video as a means for expression.  Text, audio, photographs, and other media may provide useful material for Conversational Documentary as well.  I see this sort of low-barrier-to-entry, easy mediamaking activity as part of a strategy that provides benefits both for knowledge transfer within organizations and groups, and for compiling parts of externally-facing deliverables.  

The LifeSampler

The LifeSampler is a software tool designed to address some of the difficulties faced by collaborators who wished to produce video documentation of their work.  I observed that project documentation often occured as an afterthought, resulting in incomplete capture of the process leading to a final product and rushed documentation completed near the end of a deliverable's production timetable.  I hoped that providing a simple, easy-to-use, low-barrier-to-entry video recording system might prompt users to record documentation more often during their daily work.  In order to provide motivation, I loaded the system with prompts, starting with the very basic question, "what are you doing today?" I hoped that users might engage more readily with a system that prompted a specific response than with an open-ended call to "please, just shoot some video!" For the same reason, responses are limited to sixty seconds in length, encouraging focused answers.  Users can re-record their answer as many times as they like, but cannot exceed the hard time limit.

The system is built in Java on top of the Processing API, and uses AppleScript to control RecDV, a simple video recorder, and FFMPEG, a video compression toolkit. It interfaces with the Media Repository via XML-RPC.  Due to its reliance on RecDV/applescript, the system runs on a Power Mac. I am, however, investigating better solutions (including a Flash-based interface leveraging Red5) to achieve platform independence.

The original system was situated, meaning that camera, lighting and audio setup were pre-set.  Users did not have to concern themselves with setup tasks; likewise, post-recording tasks like compressing video and uploading to a media repository -- in our case, the AME Media Repository system -- were handled automatically by scripts in the LifeSampler system.  RFID tags embedded in hand-made bracelets (created by fellow AME researcher Rebecca Stern) or other aesthetically-appealing objects allowed users to identify themselves to the system without remembering a username/password, minimizing cognitive load. Since the device is situated, users are uniquely identified, the prompt is known, and recordings are brief enough that they can be presumed to respond to the prompt relatively concisely, videos created by the LifeSampler are already well-annotated upon upload into the Media Repository.

As users began working with the LifeSampler, I added several features to respond to user demand. First, multiple questions, randomly selected, were added. Next, I allowed users to record their own prompts, as well as answers.  As the number of user-submitted questions grew, I added a list-view, where users could browse and select probes in addition to using the random probe-selection feature.  Finally, users complained that they had to come to the physically-situated kiosk in order to use the system.  In response, I created the latest version of the software, which is intended to function on any Apple computer rather than just the situated kiosk. This version sacrifices RFID login, but now pulls prompts from a central server.  This significant change in architecture will enable future development, such as tracking probes skipped, responded to, and so forth, and potentially more targeted probe selection in place of the current random mode.

Future Directions

The LifeSampler system is concerned mainly with capturing and annotating brief, focused video clips and uploading them into a media repository.  I plan to extend this work by enhancing the media repository's search and playback functions. In an ideal world, the system would respond to a query for information on a subject by piecing together not just LifeSampler submissions, but other images, documents and audio clips to produce something akin to a customized 3-minute movie that takes into account all knowledge about the intended viewer captured by the Media Repository's profile, submissions, and so forth -- what they've already expressed an interest in, where they work, who they work with, who they socialize with.