Skip to content
BRIC BRIC Docs

Data Minimization

There are many ways to collect only the data you need using REX. Web history or LLM conversations can contain particularly sensitive data, which is why data minimization is built into REX’s functionality. This foregrounds participant privacy considerations and facilitates researcher compliance with IRB and GDPR standards.

Web Browsing History

Section titled “Web Browsing History”

Levels of data collection (least to most sensitive):

  • Time stamps
  • Category
  • Domain
  • URL and title

REX will always collect browser time stamps. Researchers can collect this data alone if they are interested in questions of time spent browsing or questions related to time-of-day usage.

Researchers can use the rex-lists module to establish site categories and control for which sites only category will be collected. For example, you can block Instagram, TikTok, and Facebook and have them simply listed under the category of “social media.” This feature is especially useful for filtering data upon collection. This allows participants to know exactly what data of theirs will be included in analysis, often excluding sensitive information and facilitating privacy. This also allows researchers to be specific about data collection and management in the IRB and GDPR approval process.

Regarding domains, url, and titles, researchers can use the rex-content-processing module to make additional redactions. Common redactions include email addresses or names.

LLM Conversations

Section titled “LLM Conversations”

Levels of data collection (least to most sensitive):

  • Time stamps
  • Titles of conversations
  • Full content of conversations

Just as with the web browsing history, REX will always collect time stamps of LLM conversations. Researchers can collect this data alone if they are interested in questions of time spent interacting with an LLM or questions related to time-of-day usage.

Researchers can also use the rex-content-processing module to make additional redactions to the titles of conversations or the content of conversations.

Local exploration

Section titled “Local exploration”

Data need not necessarily leave a participant’s own machine. Data can be examined without being uploaded to a server in some cases. A local-only approach may work for an interview-based study where researchers and participants walk through data visualizations on the participant’s computer.