Viviane (Vivi) Ito

About

Hi! 👋

I am a fourth-year Ph.D. student in Information Science at the University of North Carolina at Chapel Hill and a Graduate Affiliate at CITAP. My work is currently supported by the Google PhD Fellowship.

I've always been a creative person. I remember having an assignment as a kid to craft a prototype of an invention and present it to all my colleagues at school. I crafted a 10 in 1 apparatus, that cleaned, flied, recorded sounds, worked as a remote control… I was a very imaginative little human being.

Now, I reimagine things with AI. I collect AI data and understand their perspective. I also see how people are interacting and reimagining things with AI aid.

I reimagine…

How patients and doctors can rely on better AI systems to help them with diagnosis and experience illness.
AI-generated images without stereotyping gender roles.
How AI systems can be more ethical and transparent about their outputs.

Overall, I imagine a future of AI-human collaboration with more trust and transparency.

My research interests bridge computational and social science domains. On the one hand, I study the linguistic properties of AI systems, exploring how they evolve with advancing models and technology. On the other hand, I examine how people interact with these systems, seeking to understand how they process and make sense of retrieved information.

To analyze these objects, I use natural language processing methods that I have learned in my BA and MA in Linguistics, both at the University of Sao Paulo, Brazil.

itovivi [AT] unc [DOT] edu

Projects

Women's Health Information Seeking (aug 2025 - present)

For my dissertation, I am studying how women find information about perimenopause and menopause. I am particularly interested in their use of digital platforms, including AI and how their use may change the doctor-patient relationship.

AI-brary (mar 2025 - present)

With my co-author, Lyric Grimes, we investigate how AI models generate stereotypical portrayals of librarians. We are expanding our work to a collaborative website where other researchers can upload their generated images and donwload our generated images for their use, and also working on a new approach, understanding how the images of libraries themselves is presented by text-to-image models.

LLM Audits (jan 2024 - present)

My work as a member of the Search Prompt Integrity and Learning Lab (SPILL Lab), explores how information flows through platforms in the evolving search space – permeated by AI systems and malicious agents. The lab is an interdisciplinary and transatlantic environment, with researchers coming from diverse backgrounds. Since the environment is very collaborative, I am participating in all research stages: from brainstorming research ideas to data collection, data analysis, literature reviews, and writing papers with our results. My most recent assignment is to study Wikipedia and Reddit data to understand how people make sense of online information.

Women's Biographies (aug 2023 - present)

Is the risk worth the reward? In this project, we are assessing the tension between notability and harassment. We aim to understand if notable women refrain from having a biography or photo on Wikipedia for fear of being harassed. As an RA in this space I’ve helped design survey questions, curate the survey sample, and write the data analysis plan. This survey is currently being piloted - after which I will also be responsible for analyzing the collected data.

Past Projects

Discurso Y Dolor (jul 2021- jan 2022)

In this project, funded by FONDECYT, I was supervised by Prof. Mariana Pascual from the Pontificia Universidad Católica de Chile.

My goal was to study a corpus of interviews with patients diagnosed with endometriosis. The disease causes chronic pain, and takes in average 4 to 11 years to diagnose. This delay is often due to failures in patient-doctor communication.

Questions that were the basis of the project:

What are the patterns that the patients use in discourse?
How can doctors be more precise in early diagnostics through language?
Which language resources do patients recur to express their pain?

Gender Bias In The Careers Of Olympic Athletes (jan 2021- mar 2022)

I was a Research Assistant in this project during my BA in Linguistics at USP, supervised by Prof. Marcos Lopes. It presented two central objectives, anchored in the hypothesis that the sports careers of professional athletes show mishaps directly related to gender. The main objective was to understand whether themes such as pain, career transition, and motherhood are relevant in differentiating the trajectory of female athletes.

To enable this study, computational analyses were made to quantitatively evaluate textual data, since the corpus of the work consists of a set of interviews. This methodology should allow the extraction of the most critical topics in the differentiation of interviews of men and women, besides generating metrics for visualizing such differences.

Papers

Viviane Ito and Lyric Grimes. Occupational Identities in the Age of Generative AI: The Case of Librarianship.
Anna Beers, Viviane Ito, Agustin Orozco, Pablo Aragón, Patrick Gildersleve, and Francesca Tripodi. Refractive Datasets as a Sensemaking Methodology in Closed Data Ecosystems.
(Accepted to AoIR 2025) Viviane Ito and Lyric Grimes. Adorkable AI: How Algorithms Shape Librarians Stereotypes in Brazil and the US.
(Accepted to ASA Meeting 2025) Viviane Ito. Gendered-AI: Analyzing Depictions of Household Labor.
(Accepted to ICWSM 2025) Patrick Gildersleve, Anna Beers, Viviane Ito, Agustin Orozco, and Francesca Tripodi. WikiReddit: Tracing Information and Attention Flows Between Online Platforms.
Viviane Ito and Mariana Pascual. Exploring Women’s Chronic Disease Experiences: A Mixed-Methods Analysis of Endometriosis Narratives.
(Working Paper) Tensions Around Wikipedia Notability: Is the Risk Worth the Reward? With Dr. Francesca Tripodi, Dr. Sarah Gilbert, Dr. J. Nathan Matias, and Ady Weng.

Page updated

Google Sites

Report abuse