Arnold Library is a biomedical library part of Fred Hutchinson Cancer Research Center recently migrated to a CONTENTdm system and aims to make the Archive dataset fully searchable for all Fred Hutch employees. To support the library in better facilitating information discovery for users, this project examined available CONTENTdm features and further improved the metadata standards. As a result, the project produced guideline documentation on how to employ finding aids to improve the search functionality of the collection. Additionally, this built the foundation for future work to further improve user information access and discovery within the CONTENTdm system.

The Japanese American experience has long been defined by incarceration; however, that has led many to overlook other important wartime experiences. This digital humanities project aims to highlight the role that Minnesotan colleges and other educational institutions played in defining the nisei experience during World War II. Using archival material, the project demonstrates that there was no single institutional response; rather, each institution charted its own path with varying results for nisei. By focusing on the institutional responses, this project promotes a deeper, more nuanced understanding of the ways in which institutions can either help or hinder those facing injustices.

Misinformation is a significant continuing problem in online discourse, affecting our trust of others and institutions. While unconventional, studying fandoms such as ARMY and the brony fandom yielded interesting insights into how commonplace misinformation occurs, and how these online fandoms developed strategies to counter them. This study expands on Prof Lee’s work on ARMY and applies it to the brony fandom, discovering that both fandoms share techniques but also have unique yet effective approaches to tackling misinformation. The results have significant implications for mitigating misinformation in online discourse without affecting community cohesion, restoring online trust in information.

Homelessness is a multifaceted crisis that includes many common information and community challenges, such as social instability, systemic inequity, and discrimination. We combine social network analysis, Natural Language Processing, and stochastic topic modeling to measure online information sharing among U.S. homeless care organizations. Analyzing four years of social media data from major federal homeless care communities reveals evidence of a robust-growing homeless service community interacting with the marginalized Hispanic/Latinx population shortly after the announcement of COVID-19. This study offers valuable practical policy implications for U.S. homeless services organizational learning.

The pandemic and the shift to remote work have made it more difficult for college students to meet new people and make deeper connections with those with common interests. Based on student interviews, we found that the majority of the participants did not stay in touch with their peers after the class quarter ended, and the main reasons behind this were due to the lack of initiative where it was hard to “break the ice” outside of classwork. We designed a simple application that would prompt those working in teams to find common interests and spark conversations to deepen relationships.

Classification schemes typically used in libraries are strictly hierarchical, which can be useful for organization of resources, but does not reflect the more general structure of the universe of knowledge. Epistemologically, knowledge has other types of relationships between resources (e.g. ontologies reflect this well). Graph theory in the computer science field provides compelling transformations from existing classification schemes to novel representations. Tries, the structure that decimal schemes use, when transformed to more general structures, solve a number of issues, both epistemological and logistical (logistical issues being shelving, discoverability, and other physical library considerations).

In recent years there has been an abundance of mis/disinformation around US elections on social media sites like Twitter. Research collectives such as the Election Integrity Partnership investigated this information in real time to analyze and disseminate important details across election stakeholders. However, not all researchers have the quantitative skill sets necessary to access this data at scale. Our project improves the research process of qualitative researchers by creating a Python Jupyter Notebook that allows researchers to gain insight from datasets relevant to the US election.

Academic research materials are a key resource for students and educators. Authorized actions - for example, making digital or print copies, scholarly sharing with colleagues, or disseminating works to other libraries (interlibrary loan) - are described in their legal licenses. For this project, we reviewed, amended, and categorized 698 licenses and legal materials in order to assist students, faculty, and staff in utilizing content in their academic research. With the conclusion of this project, the updated Alma database will better inform individuals on matters of Fair Use, Data Privacy, and Accessibility.

Open educational resources are teaching and learning resources that are free of cost and access barriers. Libraries play a significant role in strengthening networks for the creation, use, and distribution of OER. I worked alongside members of the Digital Library of the Caribbean to develop and distribute a survey to Caribbean Studies instructors across a variety of fields and institutions, both in the U.S. and internationally. We sought to learn about their current practices using, creating, or sharing OER, as well as their needs and priorities. The results of this survey will inform the design of OER initiatives at dLOC.

Joe, a partner at a VC shares a password with a team of 35 but paid for one seat with PitchBook. This is a violation of the terms of use, poses a security risk, and adds friction to account managers day to day. We implemented an intuitive interface built on a machine learning foundation by finding significant attributes in the user login data that will detect a password-sharing event. The valuable insights on usage data revealed strong correlations between a suspicious session with various types of usage that would help account managers monitor accounts with unusual usage on the platform.