Loading...
Visualizing bias against Asian Americans during the pandemic
View project
In 2020, Asian American communities faced a barrage of hate coinciding with the coronavirus pandemic. The founding board members of The Asian American Foundation (TAAF) hypothesized that the bias incidents and hate crimes reported in the news were the just tip of the iceberg. The team wanted to go beyond headlines and show the lived experiences of those who had been targeted as pandemic fears stoked deep-seated xenophobia across the country. At the same time, the organization had yet to hire full-time staff and didn't have the capacity to collect reports directly from the community. As the project lead, I was responsible for exploring alternative ways of tracking reports that would scale without a large physical footprint.
In our initial survey of existing efforts to track bias incidents and hate crimes across different communities, I found that the The Anti-Defamation League (ADL) had been tracking reports of anti-Semitism in the United States for over four decades. While TAAF didn't have the infrastructure to build out a network of offices, we were able to partner with subject matter experts at ADL's Center on Extremism to build a search query and identify reports of incidents and crimes reported by local news. Following ADL's practices, I defined a set of categories and tags (e.g. incident type and location) to classify the incidents we found. Our initial crawl revealed that the volume of coverage paled in comparison to figures compiled by community-based reporting channels like Stop AAPI Hate and also disproportionately reported physical assault and property-related crimes.
As the vast majority of media reporting didn't include the perspective of the person targeted, I began looking at social media platforms as an alternative source. Twitter, where messages were public by default and often included firsthand accounts, offered unique insight into how people had been personally affected by the surge in hate. After optimizing our query for the platform, the team began annotating search results (tweets) that reported bias incidents or hate crimes. After manually tagging a couple thousand tweets, I worked with engineers at a local startup to train a natural language processing system to help identify tweets reporting an incident or crime. Using a combination of machine learning and human review, the team identified over 7,000 tweets between January 2020 and December 2021.
I used Google Sheets and Apps Script to pull key insights from the data and share them internally with the board. As point of reference, my team compared the findings with community-based reporting efforts such as Stop AAPI Hate and found similar patterns across the types of locations where incidents and crimes took place as well as the gender and ethnicity of the people targeted. I worked with Pitch Interactive to build a data visualization that would communicate the impact of hate on the community to viewers who were unfamiliar with the issue. We displayed our news dataset alongside thousands of tweets to allow viewers to compare and contrast the narratives across platforms. The visualization also leveraged the tags we developed with ADL, by allowing users to group the data by incident type, location type, or the gender of the person targeted.
The visualization was shared widely, from a Washington Post event to a publicly-owned Argentine television network. @TwitterDev featured the work as a case study in their #BuildForGood developer relations program. Our work also conveyed to existing community-based organizations that TAAF was invested in studying bias incidents and hate crimes. Following the release of the visualization, my team partnered with researchers at Stop AAPI Hate, the largest incident reporting platform by volume, to draft shared standards for community-based report collection.
