This will be an afternoon workshop on the ICWSM workshop schedule. As described in the rationale, this workshop will be held in an interactive format. It will not feature a series of presentations.
The organizers are currently developing the final schedule. For the time being, here is the proposed timeline for the afternoon that was developed for the rationale.
- Introduction of organizers and participants
- Brief description of audits, the main challenges and explanation of Exercise 1
- Exercise 1: Facebook groups exercise
- Short break
- Explanation of Exercise 2
- Exercise 2: Group audit using five different strategies
- Group discussion
- Unofficial group drinks (optional)
- Unofficial group dinner (optional)
(We welcome activity suggestions from registered participants at email@example.com.)
Exercise 1: The first exercise will begin with the creation of a Facebook group consisting of the workshop participants and organizers. We will spend the first fifteen minutes of this session populating this group feed with articles and comments about algorithmic curation, awareness, application, ethics, and policy. The purpose for this is two-fold: (1) We plan to use this group throughout the workshop and perhaps after as an archive of our discussion, (2) Group feeds differ from the basic Facebook News Feed in that there is no option to view the feed in chronological order. From the onset, prioritization algorithms dictate the order of presentation for the audience. We will use this data for the remainder of Exercise 1 to create a systematic face-to-face real-time collective group approach to discovering the prioritization elements that curate this group feed.
Exercise 2: We briefly present five audit strategies in the Workshop Description above. In Exercise 2, the workshop participants will divide into five groups, one for each of the strategies. The groups will be presented with the same data set, tentatively, the Facebook group data. The goal of each group is to use their assigned audit strategy to uncover feed personalization and prioritization features. We will then compare the resulting features compiled in each of the five groups.