The internet has become the de-facto information source for individuals – from finding a new cookie recipe, to learning how a transistor works. Tools for helping users find answers to their questions have grown tremendously in response; a user can get an answer in their search results for when the next Pirates game is, or get a list of facts about their favorite movie actress. However, for many questions, such as buying a car, there isn't one right answer – the best choice for an individual depends on their particular set of circumstances and personal preference. For these situations, it's up to the user to make sense of the answer space: accumulate what options are available, what the differencess and features of these options are, and eventually choose between them.

This process, sensemaking, is a highly iterative and cyclical process, where information is constantly being found, incorporated, restructured, summarized, and generalized. As users continue to collect new information, they need to adjust and restructure their existing information, while incorporating the key known points of the new information into their understanding of the problem. This constant adjustment of both the data and structure surrounding the data puts a significant mental burden on users, and often requires them to resort to external means to track and manage this information. In the context of online sensemaking, this can be done in notepads, tabs, word processors, spreadsheets, kanban boards, or even emails. As users proceed to move along with their data in this process, they need to manually update and transfer data between these tools, which might often be more trouble than its worth.

In this work, I explore how integrated, in-situ sensemaking tools allow users to manage, structure and evaluate their sensemaking data more effectively. I posit that by reducing the interaction costs for users to externalize their mental models in four key stages of the sensemaking process: seeking, triage, structuring, and evaluation, users can more adequately juggle the large amount of information required to perform effective sensemaking. First, I explored and developed a workflow for performing sensemaking using crowdworkers with the Knowledge Accelerator System. Crowdworkers, using this workflow, were able to answer complex questions while spending less than 5 minutes on a single task, suggesting a lightweight scaffolding that could be adapted for use in individual sensemaking. This led to the creation of a sensemaking framework that connects the process of sensemaking with the cognitive processes and tools users leverage for their sensemaking tasks.

Using this framework, I then went on to develop three systems for supporting individual sensemaking: Bento, Distil and Meta. These systems were designed to provide in-situ support for one or more of the four phases of seeking, triage, structuring, and evaluation. Through these tools, I was able to identify several key strategies for creating effective online sensemaking tools. These strategies suggest several promising directions for the future of integrated, in-situ browser sensemaking tools.

145 pages

Thesis Committee:
Aniket Kittur (Chair)
Brad A. Myers
Adam Perer
Jaime Teevan (Microsoft Research)

Jodi Forlizzi, Head, Human-Computer Interaction Institute
Martial Hebert, Dean, School of Computer Science

Return to: SCS Technical Report Collection
School of Computer Science homepag e

This page maintained by reports@cs.cmu.edu