Project teams are increasingly common in classes. However, simultaneously assembling dozens of effective teams is challenging because it is difficult to take into consideration several different constraints. This thesis develops an interface that facilitates the process of forming teams within classes. It introduces Teammatic, a mixed initiative interface for instructors to work in conjunction with an algorithm. We created an algorithm that will account for several constraints, as inputted by the instructor, in order to create teams among the students. Common constraints include distributing roles and ensuring overlapping schedules amongst teams. Our results show that Teammatic produces teams that have equal or higher compatibility scores than teams created manually. Furthermore, Teammatic allows instructors to form teams given various constraints in significantly less time.

Searching and exploring online is a part of our everyday lives – shaping how we learn, work and innovate. However, today, people are drowning in information, with few mechanisms for managing or synthesizing large volumes of disparate information. It is a struggle to find the right information or identify relevant unknown unknowns for those who lack knowledge of a particular domain or well-defined goals. Even experts juggle dozens of disparate information silos spread out across different apps, websites, and work sessions. This is cognitively overwhelming and time-consuming, preventing people from developing a comprehensive understanding, gaining deep insights, and achieving their creative potential. This is especially true in complex creative information work like scientific research, founding a startup or innovating to protect the public during a pandemic.

As the Web paradigm evolves to include Generative AI models and beyond, we are experiencing a shift in how we search, learn, work and create. With this transformation in human-AI interaction, it is important to investigate how we might present the user with the right information in the right context, the right representation, and at the right time. This thesis explores this in the context of cognitively complex information work (such as knowledge discovery, synthesis, and creativity). It presents two types of contributions: (1) Empirical studies that further our understanding of how people explore, make sense of, and create using information on the Web. The studies follow a mixed-methods approach, combining large-scale and longitudinal quantitative data analysis with in-depth qualitative inquiry. (2) Computational and interaction techniques that augment these cognitive processes by seamlessly integrating knowledge from the Web into the user’s work context.Each study observes user behavior, challenges, and strategies at different stages of information exploration, sensemaking, and creative processes. Each system introduces an approach for inferring contextual signals from user-generated artifacts. For example, such as CoNotate mines an individual’s unstructured artifacts for knowledge gaps and patterns to make query suggestions, InterWeave analyzes and presents suggestions in the user’s evolving sensemaking structures to present suggestions, Relatedly mines existing knowledge structures on the web from previous users to present dynamic topic overviews, and Orchid enables users with affordances to specify and refer to personal, project-level, and external contexts. User evaluation studies demonstrate how these techniques, mining rich contextual signals from work done during cognitive processes, can promote information exploration, synthesis, and creativity. Overall, this dissertation demonstrates the potential of distilling and integrating the immense knowledge on the Web within the context of everyone’s workflows to help augment cognitively complex work.