We aim to understand how framing and reframing of real world events influence the ability of everyday people to seek and
evaluate information online. Our goal is to build NLP tools that can capture the full narrative context from a social media post,
build contextual representations of these texts and effectively address issues like bias and misinformation online and help users
better understand the context that shapes the information that they consume.
Narratives on Social Media
Our lab uses computational approaches to investigate the dynamics of storytelling on social media,
exploring why individuals share stories online and how these narratives contribute to broader social discourses.
We build models to detect narratives in various contexts and analyze the formation and evolution of narratives
through interconnected posts, focusing on their development, impact, and the strategies to counteract harmful content.
Our research aims to deepen the understanding of individual posts within larger narrative frameworks and to provide
insights into the collective construction of online stories.
Social Norms in Health Contexts
We leverage text analysis methods to study the evolution of social norms surrounding health behaviors,
particularly in the context of risky behaviors. By employing natural language processing techniques to analyze
social and entertainment media, our research seeks to capture the shifts in these norms over time and across different
communities. Our long-term goal is to leverage this knowledge to design interventions that correct misperceptions and
promote healthier behaviors.
AI for Education and Information Seeking
This area of research focuses on the use of AI to enhance education, particularly in STEM and medical fields.
We explore the creation and evaluation of pedagogical practices generated by AI, aiming to model learners'
understanding and provide tailored mediations that reinforce learning. Additionally, we are interesting in
exploring the influence of AI-driven learning on the acquisition of knowledge, comparing its effects with
traditional learning materials such as textbooks and videos, to understand and mitigate potential biases in
information seeking and comprehension when learners rely on AI tools.
Past Projects
Personal Values And Human Activities
The things that people choose to talk about are, in some ways, a reflection of what is important to them.
We explore how language use is connected to personal values, which are, in turn, connected to what people
choose to do. Using data from free-response surveys and social media, we build models and lexical resources
for the measurement of personal values in text.
Sarcasm, Humor And Slang In Online Communication
The text that people write online in informal settings often contains features that cause problems to traditional
NLP pipelines, such as sarcasm, humor, and slang. We are working to develop new datasets, methods, and tools
that enable better semantic representations of this type of text to both improve our NLP models and to better
understand these phenomena and the people that use them in their writing.
Online Harms and Misinformation
We are focused on applying our NLP methods to helping to make the internet a safer place. This includes using
computational approaches to detecting, characterizing, and combating offensive language, bias, and misinformation
that is common in online text-based datasets.