Jordan Huffaker

I build collaborative human-AI systems that can understand and process social references —language that invokes connotations by overlaying parts of social and cultural contexts. My systems combine peoples' understanding of social and cultural contexts with the consistency and scale of computation to form powerful collective intelligence. Thus far, I have leveraged this approach to combat harmful rhetoric (e.g., the rhetoric of hate speech, harassment, and misinformation) and to improve the quality of human-AI translation teams.

I have many years of both academic and industry experience, interning at various companies throughout my undergrad and grad education. My most recent industry experience at Lilt involves conducting research to improve human-AI translation teams by studying professional freelance translators' use of neural machine translation. My work in this position has led to the development of a method called guided context building that directs human labor within human-AI teams to areas where it is most needed, improving the efficiency and effectiveness of the team.

I am currently an HCI researcher at the University of Michigan aiming to finish my PhD in June 2023. During this time, I have primarily developed two systems that aim to combat the rhetoric of hate speech, harassment, and misinformation. The first of these includes a system that can disentangle emotionally manipulative language (e.g., fear-inducing anti-vaccine propaganda) from intrinsically emotional content (e.g., a parent losing a child) through non-expert crowdsourcing. Additionally, my thesis work involves building a system using a crowdsourced panel of people to expose codified narratives within articles.