About

Cassidy R. Sugimoto is a faculty member associated with the Jimmy and Rosalynn Carter School of Public Policy at Georgia Tech. Her research focuses on science, technology, and innovation policy, as well as the ethics and philosophy of science and technology. She is involved in various research centers, including the Science and Engineering Organizations, Education, Careers, and Workforce center, and the Policy Process, Leadership, and Pre-Law program. Her academic work includes contributions to understanding the intersection of science and society, with particular attention to policy implications and ethical considerations. She advises Ph.D. students and is actively engaged in research related to science communication, information science, and the societal impacts of technological advancements.

Research topics

• Sociology
• Mathematics
• Mathematical analysis

Selected publications

• Intersectional biases in narratives produced by open-ended prompting of generative language models

  Nature Communications · 2026-01-08 · 2 citations

  preprintOpen access

  The rapid deployment of generative language models has raised concerns about social biases affecting the well-being of diverse consumers. The extant literature on generative language models has primarily examined bias via explicit identity prompting. However, prior research on bias in language-based technology platforms has shown that discrimination can occur even when identity terms are not specified explicitly. Here, we advance studies of generative language model bias by considering a broader set of natural use cases via open-ended prompting, which we refer to as a laissez-faire environment. In this setting, we find that across 500,000 observations, generated outputs from the base models of five publicly available language models (ChatGPT 3.5, ChatGPT 4, Claude 2.0, Llama 2, and PaLM 2) are more likely to omit characters with minoritized race, gender, and/or sexual orientation identities compared to reported levels in the U.S. Census, or relegate them to subordinated roles as opposed to dominant ones. We also document patterns of stereotyping across language model–generated outputs with the potential to disproportionately affect minoritized individuals. Our findings highlight the urgent need for regulations to ensure responsible innovation while protecting consumers from potential harms caused by language models. People use generative language models to inform all aspects of their lives. Here, the authors examine synthetic narratives produced by open-ended prompting of generative language models and identify patterns of omission, subordination, and stereotyping against minoritized intersectional subgroups.

  PublisherOA PDFDOI

• Language Models Generate Widespread Intersectional Biases in Narratives of Learning, Labor, and Love

  2025-10-29

  article

  The rapid deployment of generative language models (LMs) has raised concerns about social biases affecting the well-being of diverse consumers. The extant literature on generative LMs has primarily examined bias via explicit identity prompting. However, prior research on bias in language-based technology platforms has shown that discrimination can occur even when identity terms are not specified explicitly. Here, we advance studies of generative LM bias by considering a broader set of natural use cases via open-ended prompting, what we refer to as a laissez-faire environment. In this setting, we find that across 500,000 observations, generated outputs from the base models of five publicly available LMs (ChatGPT3.5, ChatGPT4, Claude2.0, Llama2, and PaLM2) are hundreds to thousands of times more likely to omit or subordinate characters with minoritized race, gender, and/or sexual orientation identities. We also document patterns of stereotyping across LM-generated outputs with the potential to disproportionately affect minoritized individuals. Our findings highlight the urgent need for regulations to ensure responsible innovation while protecting consumers from potential harms caused by language models as well as further investments in critical artificial intelligence education programs tailored towards empowering diverse consumers.

  PublisherDOI

• Impact of Marriage on Productivity and Career of Women Scholars

  2025-07-10

  articleOpen accessSenior author

  Marriage has potential impact on scholarship, especially for women, but lack of appropriate data has prevented its clear assessment. In this article we quantify the impact of marriage on women’s scholarship using open data from ORCID (23057 married women scholars are recognized), including longitudinal productivity data and career path. So far we have find marriage have short term negative impact but long term active impact on productivity of women scholars and this impact varies according to the field they worked in. The short term negative impact is more significant if they get married after starting their careers. While we continue to investigate other aspects of this topic, such as the impact of marriage on career progression, we believe this research will offer valuable insights for academic institutions and policymakers, helping to ensure that marriage does not become an insurmountable barrier to women’s academic success.

  PublisherDOI

• Author response: Gender differences in submission behavior exacerbate publication disparities in elite journals

  2025-09-09

  peer-reviewOpen access

  Women are particularly underrepresented as leading authors of papers in journals of the highest impact factor, with substantial consequences for their careers. While a large body of research has focused on the outcome and the process of peer review, fewer articles have explicitly focused on gendered submission behavior and the explanations for these differences. In our study of nearly five thousand active authors, we find that women are less likely to report having submitted papers to journals of the highest impact (e.g., Science, Nature, or PNAS) andto submit fewer manuscripts, on average, than men when they do submit. Women were more likely to indicate that they did not submit their papers (in general and their subsequently most cited papers) to high-impact journals because they were advised not to. In the aggregate, no statistically significant difference was observed between men and women in how they rated the quality of their work. Nevertheless, regardless of discipline, women were more likely than men to indicate that their “work was not ground-breaking or sufficiently novel” as a rationale for not submitting to one of the listed prestigious journals. Men were more likely than women to indicate that the “work would fit better in a more specialized journal.” We discuss the implications of these findings and interventions that can serve to mitigate the disparities caused by gendered differences in submission behavior.Publishing in high-impact scholarly journals has a significant effect on researchers’ careers. Our findings identify factors that affect submission to Science, Nature, and the Proceedings of the National Academy of Sciences of the United States of America (PNAS) and explore whether there is a relationship between gender and desk rejections or submission rates. We found no relationship between gender and reported desk rejection and a relationship between gender and reported submissions, with men having a greater number of submissions. Women were more likely than men to indicate that their “work was not ground-breaking or sufficiently novel” for the listed prestigious journals and that they were advised against submitting to these venues. Men were more likely to indicate that the “work would fit better in a more specialized journal.”

  PublisherDOI

• Author response: Gender differences in submission behavior exacerbate publication disparities in elite journals

  2025-07-14

  peer-reviewOpen access

  Women are particularly underrepresented as leading authors of papers in journals of the highest impact factor, with substantial consequences for their careers. While a large body of research has focused on the outcome and the process of peer review, fewer articles have explicitly focused on gendered submission behavior and the explanations for these differences. In our study of nearly five thousand active authors, we find that women are less likely to report having submitted papers to journals of the highest impact (e.g., Science, Nature, or PNAS) andto submit fewer manuscripts, on average, than men when they do submit. Women were more likely to indicate that they did not submit their papers (in general and their subsequently most cited papers) to high-impact journals because they were advised not to. In the aggregate, no statistically significant difference was observed between men and women in how they rated the quality of their work. Nevertheless, regardless of discipline, women were more likely than men to indicate that their “work was not ground-breaking or sufficiently novel” as a rationale for not submitting to one of the listed prestigious journals. Men were more likely than women to indicate that the “work would fit better in a more specialized journal.” We discuss the implications of these findings and interventions that can serve to mitigate the disparities caused by gendered differences in submission behavior.Publishing in high-impact scholarly journals has a significant effect on researchers’ careers. Our findings identify factors that affect submission to Science, Nature, and the Proceedings of the National Academy of Sciences of the United States of America (PNAS) and explore whether there is a relationship between gender and desk rejections or submission rates. We found no relationship between gender and reported desk rejection and a relationship between gender and reported submissions, with men having a greater number of submissions. Women were more likely than men to indicate that their “work was not ground-breaking or sufficiently novel” for the listed prestigious journals and that they were advised against submitting to these venues. Men were more likely to indicate that the “work would fit better in a more specialized journal.”

  PublisherDOI

• Gender differences in submission behavior exacerbate publication disparities in elite journals

  eLife · 2025-07-14

  preprintOpen access

  Abstract Women are particularly underrepresented as leading authors of papers in journals of the highest impact factor, with substantial consequences for their careers. While a large body of research has focused on the outcome and the process of peer review, fewer articles have explicitly focused on gendered submission behavior and the explanations for these differences. In our study of nearly five thousand active authors, we find that women are less likely to report having submitted papers to journals of the highest impact (e.g., Science, Nature, or PNAS) andto submit fewer manuscripts, on average, than men when they do submit. Women were more likely to indicate that they did not submit their papers (in general and their subsequently most cited papers) to high-impact journals because they were advised not to. In the aggregate, no statistically significant difference was observed between men and women in how they rated the quality of their work. Nevertheless, regardless of discipline, women were more likely than men to indicate that their “work was not ground-breaking or sufficiently novel” as a rationale for not submitting to one of the listed prestigious journals. Men were more likely than women to indicate that the “work would fit better in a more specialized journal.” We discuss the implications of these findings and interventions that can serve to mitigate the disparities caused by gendered differences in submission behavior.

  PublisherDOI

• A static research enterprise decouples from changes in the burden of disease

  Research Square · 2025-03-25

  preprintOpen access
  PublisherOA PDFDOI

• Gender Differences in Beginning Academic Careers and Navigating Parenthood

  2025-09-13

  preprintOpen access

  Balancing academic training and parenthood poses unique challenges, particularly during the Ph.D. stage. Using global survey data, we reconstruct life-course trajectories of academic parents to investigate gender differences in the timing of Ph.D. completion and parenthood, and their implications for careers. We identify five distinct trajectories. Women more often prioritized Ph.D. completion before age 30 and transitioned to parenthood after 35, whereas men tended to become parents during doctoral studies, often with three or more children by age 40. Early parenthood was associated with the largest gender gaps in long-term scientific citations, favoring men; these disparities disappeared among those who became parents after age 35. Men were also more likely to secure and remain in academic positions regardless of parenthood timing. These findings highlight persistent challenges for early-career mothers and underscore the need for institutional policies to better support parenting responsibilities alongside career progression.

  PublisherOA PDFDOI

• Gender differences in submission behavior exacerbate publication disparities in elite journals

  eLife · 2025-09-09

  preprintOpen access

  Abstract Women are particularly underrepresented as leading authors of papers in journals of the highest impact factor, with substantial consequences for their careers. While a large body of research has focused on the outcome and the process of peer review, fewer articles have explicitly focused on gendered submission behavior and the explanations for these differences. In our study of nearly five thousand active authors, we find that women are less likely to report having submitted papers to journals of the highest impact (e.g., Science, Nature, or PNAS) andto submit fewer manuscripts, on average, than men when they do submit. Women were more likely to indicate that they did not submit their papers (in general and their subsequently most cited papers) to high-impact journals because they were advised not to. In the aggregate, no statistically significant difference was observed between men and women in how they rated the quality of their work. Nevertheless, regardless of discipline, women were more likely than men to indicate that their “work was not ground-breaking or sufficiently novel” as a rationale for not submitting to one of the listed prestigious journals. Men were more likely than women to indicate that the “work would fit better in a more specialized journal.” We discuss the implications of these findings and interventions that can serve to mitigate the disparities caused by gendered differences in submission behavior.

  PublisherDOI

• Laissez-Faire Harms: Algorithmic Biases in Generative Language Models (Extended Abstract)

  Proceedings of the AAAI/ACM Conference on AI Ethics and Society · 2025-10-15 · 4 citations

  articleOpen access

  The widespread deployment of generative language models (LMs) is raising concerns about societal harms. Despite this, studies of bias in generative LMs, including attempted self-audits by LM developers, have thus far been conducted in limited contexts. To address this gap, this study examines representational harms in synthetic texts produced by leading language models in response to open-ended creative writing prompts based in the United States. We conduct our investigation on 500,000 synthetic texts generated by five publicly available generative language models: ChatGPT 3.5 and ChatGPT 4 (developed by OpenAI), Llama 2 (Meta), PaLM 2 (Google), and Claude 2.0 (Anthropic). We base our selection of models on both the sizable amount of funding wielded by these companies and their investors (on the order of tens of billions in USD), as well as the prominent policy roles that each company has played on the federal level. At the time of data collection (from August 16th to November 7th, 2023), the selected models were considered state-of-the-art for each company. Creative writing prompts reflect three domains of life set in the United States: classroom interactions (“Learning”), the workplace (“Labor”), and interpersonal relationships (“Love”). Informed by intersectionality theory, we considered the role of power embedded in language by creating one power-neutral scenario and one power-laden scenario for each prompt. For example, power-neutral Learning prompts consist of a single student excelling in an academic subject, whereas the power-laden prompts consist of one star student helping a struggling student in an academic subject. We then analyze the resulting model responses for textual cues shown to exacerbate socio-psychological harms for minoritized individuals by race, gender, and sexual orientation. To do this at scale, we fine-tuned a coreference resolution model (gpt3.5-turbo) to perform automated extraction of characters’ gender references and names at high precision. To evaluate our model, we hand-label the inferred gender (based on gender references) and name on an evaluation set of 4,600 uniformly down-sampled story generations from all five models (0.0063, 95% CI). Fine-tuning our model on a non-overlapping set of 150 training examples yields precision above 98% for both gender references and names. Recall rates reach 97% for gender references and exceed 99% for names. Following previous studies, we infer racial signals from first names using fractionalized counting over the Florida Voter Registration Dataset (which consists of 27 million named individuals and self-identified racial identities). We find that when LMs are used for story writing, they generate texts that reinforce discrimination against minoritized groups by race, gender, and sexual orientation. Using mixed-methods analyses, we identify three specific harms: omission, subordination, and stereotyping. Stories produced by language models simultaneously underrepresent minoritized individuals as main characters while overrepresenting them as subordinated characters. Diverse consumers, if they are to be represented at all, disproportionately see themselves portrayed by language models as “struggling students” (as opposed to “star students”), “patients” or “defendants” (as opposed to “doctors” or “lawyers”), and a friend or romantic partner who is more likely to borrow money or do the chores for someone else. The magnitude of bias far exceeds the level of "real-world" inequities. Underrepresentation of non-dominant identities in power-neutral stories exceeds national demographics in the US by up to two orders of magnitude. Meanwhile, non-dominant character identities are up to thousands of times more likely to appear as subordinated than empowered. For example, Claude casts the name ”Juan” as a struggling student 1,380 times, yet only once as a star student. We find that these harms impact every non-dominant group we studied (in the US context). These include individuals with intersectional Asian, Black, Indigenous, Latine, NH/PI, MENA, Female, Non-binary, and Queer identities. Language models propagate a plethora of stereotypes that are known to inflict psychological harm and negative self-perception, including the ” glass/bamboo ceiling”, ” perpetual foreigner”, ”noble savage”, ”white savior”, and others.

  PublisherOA PDFDOI

Recent grants

• Cascades, islands, or streams? Time, topic, and scholarly activities in humanities and social science research

  NSF · $127k · 2012–2014

• Incubators of Knowledge: Predicting protege productivity and impact in the social sciences

  NSF · $240k · 2012–2014

Frequent coauthors

• Vincent Larivière

  240 shared

• Chaoqun Ni

  Fudan University

  40 shared

• Cameron Neylon

  Curtin University

  30 shared

• Rodrigo Costas

  30 shared

• Dakota Murray

  30 shared

• Zaida Chinchilla‐Rodríguez

  Consejo Superior de Investigaciones Científicas

  30 shared

• Nicolás Robinson‐García

  29 shared

• Chun‐Kai Huang

  Curtin University

  29 shared

Labs

• Cassidy R. Sugimoto LabPI

Education

• Ph.D., Public Policy

  Georgia Institute of Technology

  2008

• M.S., Information Science

  University of Illinois at Urbana-Champaign

  2001

• B.A., Information and Computer Science

  University of California, Berkeley

  1998

Awards & honors

• Indiana University Trustees Teaching award (2014)
• National service award from the Association for Information…
• Bicentennial Award for service from Indiana University (2020…
• Committee on Institutional Cooperation Academic Leadership P…
• James M. Cretsos Leadership Award, ASIS&T (November 2009)