TIMELINE
Oct 2021 - Jan 2023
ROLE
Project Lead
CLIENTS
USC Annenberg Inclusion Initiative
LOCATION
Los Angeles
Project Overview
In collaboration with the Media Neuroscience Lab (UCSB), the Annenberg Inclusion Initiative (USC), and Snap Inc., this project aimed to assess the inclusivity of content featured on Snapchat’s Discover platform. The central question guiding our research was: How inclusive is Discover content when it comes to gender and race/ethnicity? The project represents a broader mission to support diverse, equitable, and inclusive storytelling across digital media.
The teams used an extensively tested, hybrid computational approach, in which (AII) human assessments procedures are intelligently combined with (MNL) state-of-the-art computational algorithms from computer vision and natural language understanding. Our initial analysis focused on gender and racial/ethnic representation, with plans to expand to language and topic patterns, including indicators of sexual orientation and disability. Deliverables included a detailed final report with visualizations, offering Snap data-backed insights to support more inclusive content practices across Discover channels.
My Role
I led the design of the overall workflow and played a key role in shaping the systems and tools used throughout. I designed and tested the initial pipeline for analyzing Snapchat Discover content, helping bridge computational and human-coded methods between the Media Neuroscience Lab and Annenberg Inclusion Initiative. I led pilot testing, created a scalable platform for student researchers to use in future phases, and ensured that data could be transferred seamlessly across teams. I also supported onboarding and training of new researchers, resolved workflow issues, and helped design and deliver final analyses and visual reporting for Snap stakeholders.
Preliminary Testing
I designed the workflow and pilot test of a small selection of Snap content. This process validated three primary objectives.
Design Goals
Based on research and feedback, I determined the following user and system goals:
Designing the Training, Data Collection, and Data Validation Processes
Below demonstrates the detailed user flow throughout the audit process which led to the successful collection of DEI data from 1,400+ total videos.
