Thursday, Nov. 12, the College of William and Mary’s School of Computing, Data Sciences and Physics hosted technology entrepreneur Annie Brown ’10. The event was part of the College’s Class of 1975 Speaker Series, which features leaders and innovators.
Brown majored in gender, sexuality and women’s studies and history at the College and is the founder and CEO of Reliabl, an artificial intelligence company that develops tools designed to improve online experiences for marginalized communities.
She began by detailing how her liberal arts education shapes her work. While taking an introduction to women’s studies class, Brown realized her understanding of womanhood stemmed from a male lens. She said this inspired her to create Lips, a campus print zine featuring art and poetry by women and the LGBTQ+ community.
“A lot of us had come from Virginia schools that were abstinence education, and so a lot of us were very awkward when it came to talking about [sex and sexuality],” she said. “I also felt that there were a lot of students on campus that were questioning their sexuality and didn’t really have an outlet to express that.”
She continued this work through a study abroad program in India. There, she interviewed members of the Humsafar Trust, an LGBTQ+ community group, and collected qualitative data regarding HIV prevention work. Brown said she began asking herself how she could use this research and data to make an impact.
After graduating from the College, Brown transformed Lips into a social media platform. This was around the time AI was used for content moderation, imposing bans on some LGBTQ+ and women’s health accounts.
Brown needed AI assistance for her growing platform, so she joined the University of California San Diego as an entrepreneur in residence and an AI researcher.
“What my focus was there is understanding why content moderation systems fail the LGBTQ community, why they fail women, why they fail the BIPOC community,” she said. “And what I found was that it was at the data training layer. 73% of LGBTQIA+ content was wrongly flagged as inappropriate.”
Brown founded Reliabl based on that research and the initial machine learning training system she developed for Lips.
“We’re asking AI already to make life-changing decisions, high-stakes decisions, a lot of times decisions for people that have no say in the system,” she said. “But AI is not capable of doing that because it doesn’t understand context.”
Brown said Reliabl unites users and experts within its machine learning system, providing context and mapping out the problem space. She added that an ongoing issue of machine learning models is that a homogeneous group of data scientists often trains them.
“You might be building an AI system for housing rights in San Francisco affecting Black populations,” she said. “But you are a very well-paid white male data scientist living in the most expensive neighborhood in San Francisco, so you don’t have the context to understand the whole problem.”
Reliabl strives to mitigate this issue through connections with an array of nonprofits. She mentioned Reliabl’s current projects, including a contact recommendation system for the wellness company Communia. The company is also working to counter misogyny embedded in social media platforms, which Brown said are forcing more men toward the far right.
Through another project, Reliabl aims to assist climate-vulnerable communities by mapping gaps in emergency routes and hospital access in low-income areas that are vulnerable to fire and drought.
“When you build for people at the margins, you catch all of these entities as you understand where tools fail, and you just get a better product at the end of the day,” she said.
Robert Ressler ’10, operations director for Reliabl and a research consultant for Insevia Policy Partners, discussed what he feels differentiates Reliabl.
“One of the things that Reliabl does is that it’s purposely designed to have the different data annotators be able to interact with each other, whereas these legacy [AI companies] have perfectly designed a system that is isolated,” he said.
Sarah Venieris ’28, an attendee, emphasized Reliabl’s inclusivity.
“I think a lot of the drawbacks [of AI] have been how biases have affected users, so I think that she [Brown] brings a really unique perspective to the AI sphere that’s very vital,” Venieris said.
