UX Research

Harvard Public Domain Corpus

I led an 8-week research study on the Corpus, delivering insights to drive AI development for researchers across the world

Domain

Service Design

Duration

8 weeks

Team

2 UXR Interns

Role

As the project lead, I led the study from recruitment to insight development to drive Corpus utilization

Users

Academic researchers and engineers developing AI models for research and education

Problem

The Corpus publicly launched in 2025, but the team didn't have access to data that told them about the tool's impact

Harvard Library logo, featuring the Harvard shield crest beside the wordmark.

CONTEXT

The Corpus is a dataset of 1M books for academic researchers and engineers so that they can search for data to train their AI models

Use of the Corpus must be non-profit, educational, or research-driven. Users must request access to use the Corpus

Diagram comparing two timelines across the Software Development Lifecycle, showing accessibility currently surfacing during testing versus a future state where it's addressed during design.

THE PROBLEM

The Corpus is hosted on an external tool that publicly launched in 2025, but the team didn't have access to data that told them about the tool's impact or performance

I led research into how users use the Corpus to identify opportunities for improvement, owning recruitment, data collection, and analysis

Diagram comparing two timelines across the Software Development Lifecycle, showing accessibility currently surfacing during testing versus a future state where it's addressed during design.

OBJECTIVE

Understand how individuals are using corpus data in their work to identify opportunities for improvement

The team wanted to address these questions:

  1. How do users use Public Domain Corpus Data?

  2. What makes it hard or easy to use the corpus data in their work?

  3. What outcomes or outputs has this corpus enabled?

IMPACT

At a glance

At the end of my internship, I pitched my recommendations to drive Corpus utilization for expert and non-expert users, securing approval from the service team and UX & Discovery Director

4

approved recommendations

80+

data points across interviews and survey

HYPOTHESIS

Stakeholders hypothesized that non-expert users, or those users driven by self-exploration, experience high friction because the Corpus was intended for technical use among those who didn't need much help

45%

of users were interested in self-exploration & reading

29%

of users were interested in LLM training

25%

of users were interested in academic research

Access requests, n =211

GATHERING DATA

Methodology

Phase 1: Survey

I sent a survey to all 211 users who requested access to the Corpus both as a way to gather quantitative data and recruit respondents for interviews. Survey data revealed satisfaction levels and trends in user engagement

Sample survey questions

Phase 2: Recruiting interview participants from survey respondents

From 15 survey respondents, I recruited 2 users along with 2 stakeholders for Corpus interviews

Participant tracker
IDRoleUsed Corpus Data?Perspective they giveInterviewer
P1HobbyistNoNon-technical userKelly
P2FacultyYesTechnical userAlly
S1Project LiasonN/ASolution constrainstsAlly
S2Service OwnerN/ARequirementsAlly

Phase 3: Stakeholder interviews

Stakeholder insights revealed known issues and what kinds of recommended improvements for the Corpus are feasible to implement

Phase 4: User interviews

2 user interviews provided qualitative insights into their workflows and where friction occurs in their journey

Sample interview questions

CONSTRAINTS

Navigating contrainsts

  1. Pivoting to conducting stakeholder interviews due to small sample size

From an external pool of potential participants, 2 users signed up for interviews. I conducted stakeholder interviews to move the project along and learn more about known issues

Diagram comparing two timelines across the Software Development Lifecycle, showing accessibility currently surfacing during testing versus a future state where it's addressed during design.

  1. Since the Corpus is hosted on an external tool, feasibility was limited to how the data is structured rather than how the platform is designed

Globus tool

Cloud-based service platform that hosts Corpus, designed to move many large files

Dataset documentation

Additional types of info about the Corpus available to users with access

Library.Harvard homepage

Public-facing webpage where users can learn about the tool and request access

Insight 1

Survey and interview data validated the hypothesis that those driven by self-exploration experience high friction or quit using the Corpus

47%

of survey respondents have not used Corpus data

of survey respondents have not used Corpus data

33%

of survey respondents have used Corpus data

of survey respondents have used Corpus data

20%

of survey respondents downloaded but not used Corpus data

of survey respondents downloaded but not used Corpus data

Finding

Help was inadequate for those who were missing technical skills, storage, or tools for data management

Recommendation

Bridge technical knowledge gaps among non-expert users for easier use with links to resources, FAQs, and a persistent contact point

Insight 2

Most users abandoned the Corpus upon initial exploration of the data because they couldn't tell if the data was good enough to use and they didn't have the time or resources to keep digging

66%

of survey respondents stopped engaging after briefly exploring the data

of survey respondents stopped engaging after briefly exploring the data

33%

of survey respondents stopped engaging before exploring the data

of survey respondents stopped engaging before exploring the data

Finding

Lack of expected datasheet had users question data quality

Recommendation

Set expectations for data quality and technical requirements (e.g. computing capabilities) on the Library.Harvard page

Insight 3

Finding specific files was very difficult even for expert users because naming matched the system rather than user logic

"I haven't published anything yet because building the pipeline to extract the metadata to locate books is taking a long time”

Expert user

Finding

Filtering and locating specific books were challenging. Both expert and non-expert users found folder and file names vague

Recommendation

Orient users to the Corpus with documentation on the file structure of the Corpus with with basic details about folders

what these insights reveal

Users abandon the Corpus because they don't know what to expect, where to find the data they need, or how to get help

Impact

I pitched my recommendations to drive Corpus utilization for non-expert users, securing approval from the team and UX & Discovery Director

Both expert and non-expert users run into friction using Corpus. But focusing on the needs of non-expert users will also streamline expert workflows

Set expectations

surface data quality and technical requirements for transparency

Orient users

document the Corpus's file structure so users can find what they're looking for

Make help easy to find

FAQs, video tutorials, or a persistent contact point for support

Takeaways

  1. If I were to continue this project, I would keep recruiting with targeted recruitment with Harvard affiliates until I've reached enough participants

  2. I learned how to design a survey to still get insights in the case that we do not hear back from many participants

  3. I learned to bring in stakeholders early and often throughout the research process to validate direction