HCII PhD Thesis Defense: Andrew Kuznetsov
When
-
Where
Newell-Simon Hall 4305
Description
Thesis Title: AI-Mediated Distributed Sensemaking
Meeting ID: 987 7220 7098
Passcode: 581860
Committee:
Aniket Kittur (Chair)
Anita Woolley (Chair)
Kenneth Koedinger (Internal Member)
Ryen White (External Member)
Abstract:
Although the research community has proposed many systems to help people collect, organize, and annotate snippets of information, a gap remains between tools that support individual sensemaking (e.g. figuring out a software bug) and the vision of distributed sensemaking-where individuals can interact with the sensemaking artifacts of others (e.g. partial list of priorities from a meeting, an old query, a coworkers debugging session). This vision has a newfound urgency in today's world: recent advances in AI copilots and agents have unexpectedly made individual work collaborative, and now more than ever-no job is a one person job.
Consequently, there is a need to reimagine the work we do. Distributed sensemaking, previously often a field of study of specialized teams in healthcare, firefighting, and military command-finds itself relevant in our newly-collaborative everyday knowledge work-in our browsers, trip plans, and codebases. We must now face the challenge of sharing and absorbing large amounts of task context well beyond typical crowdsourcing micro-tasks such as image labeling. To keep our new coworkers informed, we need to support sharing our perception of that massive context at many levels, e.g. our specific judgements, general opinions-or risk leaving our "intelligent" assistants behind. Previously independent tasks, such as spreadsheet editing, now resemble the work of a people manager or organizational leader, soliciting and integrating input from a team of others. Proverbially: We need a new way of cooking if we're adding more cooks to our kitchen.
This dissertation outlines a path forward by introducing the framework THESEUS (Towards Helpful and Effective Sensemaking: Enabling User-Centric Synthesis) which focuses on capturing and applying provenance metadata during the (1) initial collection and organization, (2) interpretation, and (3) contextualization and integration of a piece of task information. Through prototypes and empirical studies, I investigate how technical systems can augment each process in-context, building an inductive foundation for supporting sensemaking in a distributed fashion. Finally, I discuss the AI-mediation of these sequential processes, outlining a future vision for end-to-end AI-Mediated Distributed Sensemaking. Dissertation Folder