Queer in AI Workshop @ NAACL 2025
May 4, 2025
In-Person (Navajo/Nambe) and Virtually
Queer in AI is organizing a hybrid (with virtual attendance) workshop at NAACL 2025. We want to bring together researchers and practitioners working at the intersection of linguistics, queerness, and natural language processing to present their work and discuss these issues. Additionally, we will provide a casual, safe and inclusive space for queer folks to network and socialize. We will have in-person and virtual components, so regardless of your physical location, we hope that you will be able to join us as we create a community space where attendees can learn and grow by connecting with each other, bonding over shared experiences, and learning from each individual’s unique insights into NLP/CL, queerness, and beyond!
You need to be registered for at least the workshop sessions (either in-person or virtual) to be able to attend the workshop. We regret that we are unable to offer financial assistance for registration.
Code of Conduct
Please read the Queer in AI code of conduct, which will be strictly followed at all times. Recording (screen recording or screenshots) is prohibited. All participants are expected to maintain the confidentiality of other participants. NAACL 2025 adheres to the ACL Code of Conduct and Queer in AI adheres to Queer in AI Anti-harassment policy. Any participant who experiences harassment or hostile behavior may contact the ACL exec team, or contact the Queer in AI Safety Team. Please be assured that if you approach us, your concerns will be kept in strict confidence, and we will consult with you on any actions taken.
Provisional Schedule (in MDT/local time)
9:30 - 10:30 Introduction
10:00 - 11:00 Contributed Talks - In Person (name of presenter in italics):
Dehumanization of LGBTQ+ Groups in Sexual Interactions with ChatGPT (Alexandria Leto, Juan Vásquez, Alexis Palmer, Maria Leonor Pacheco)
Studying the Representation of the LGBTQ+ Community in RuPaul’s Drag Race with LLM-Based Topic Modeling (Mika Hämäläinen)
Guardrails, not Guidance: Understanding Responses to LGBTQ+ Language in Large Language Models (Joshua Robert Tint)
Do language models practice what they preach? Examining language ideologies about gendered language reform encoded in LLMs (Julia Watson)
11:00 - 11:30 Contributed Talks - Virtual (name of presenter in italics):
Leveraging Large Language Models in Detecting Anti-LGBTQIA+ User-generated Texts (Quoc-Toan Nguyen, Josh Nguyen, Van-Tuan Pham, William John Teahan)
A Bayesian account of pronoun and neopronoun acquisition (Cassandra L Jacobs, Morgan Grobol)
11:30 - 12:00 Sponsor Presentation
12:00 - 13:30 Lunch Break
13:30 - 14:30 Keynote on Processing Trans Languaging (Vagrant Gautam)
Trans languaging and trans languagers are linguistically fascinating because we complicate theories of what words mean and how we actually use them. Traditional understandings of reference and agreement break down with phenomena such as multiple pronoun use (e.g., 'she/they's), mixed-gender reference (e.g., 'they/them girlfriends') and multi-entity reference (e.g., being out in selective contexts). Since these are also often long-tail phenomena that are typically understood with additional context, this makes our languaging particularly hard to process computationally. In this talk I will show how trans languaging offers a real-world testbed for broader challenges in NLP research, such as conceptualization and operationalization of abstract concepts, generalization on long-tail phenomena, processing language independent of context, and ethical issues like surveillance, representational harms and quality-of-service differentials.
14:30 - 16:30 Keynote/Fireside Chat on Code-Switching and Indigenous Language Revitalization in New Mexico (Rebecca Pattichis, Emery Sutherland, Kyran Romero)
It is no secret that mainstream NLP methods are often incompatible with real-world language use by relying on infeasible resources (e.g., large ‘clean’ datasets, several GPUs) and unrealistic operationalizations of language (e.g, parallel monolingualism). It is critical, then, to reconceptualize and repurpose NLP for our own contexts. In this talk, we will Navajo Automatic Speech Recognition, analyzing token levels for code-switching metrics using Northern New Mexican Spanish and English, and conclude with a reflection on restructuring the foundation of our Indigenous technical tools to best serve our communities. After the presentation, we will open to a fireside chat where the audience can engage in our broader discussion.
Speakers
Vagrant Gautam is a PhD candidate at Saarland University and the incoming 2025 Independent Postdoctoral Fellow at the Heidelberg Institute for Theoretical Studies. Their interdisciplinary research is broadly about social and technical aspects of trustworthy NLP – especially fairness, reasoning and robustness. On the technical side, they propose methods for and evaluations of these aspects inspired by linguistics. On the social side, they investigate gaps and opportunities in how we define and operationalize abstract concepts like ‘gender’, 'bias', etc., as well their societal implications.
Emery Sutherland is a member of the Navajo Nation and is a master's student at the University of New Mexico, specializing in Electrical Engineering with a focus on Signal Processing. Previously, he worked on rover projects, developing localization techniques for search operations. Currently, he is using modern AI methods to train computers to translate and speak the Navajo language.
Kyran Romero is a member of the Jemez Pueblo tribe and a senior undergraduate student at Stanford University. Studying Human Centered AI under the Symbolic Systems program, Kyran hopes to find innovative ways to ensure technology creation meets the needs of Indigenous communities. Creating tools that make effective change within his community while respecting their values and traditions is the forefront of Kyran’s research.
Rebecca Pattichis is a Data Analyst for UCLA’s Engineering Pathway Program. She prioritizes increasing the capacity of language communities in developing their own language technologies. As a Research Associate at the CYENS Centre for Excellence, she helped kickstart a team of Cypriot researchers developing NLP tools for Cypriot Greek. As a Google DeepMind Fellow at the University of California, Los Angeles, she adapted code-switching metrics to the prosodic unit utilizing Northern New Mexican Spanish and English transcripts.
Accepted Submissions
Archival
Studying the Representation of the LGBTQ+ Community in RuPaul’s Drag Race with LLM-Based Topic Modeling (Mika Hämäläinen)
This study investigates the representation of LGBTQ+ community in the widely acclaimed reality television series RuPaul’s Drag Race through a novel application of large language model (LLM)-based topic modeling. By analyzing subtitles from seasons 1 to 16, the research identifies a spectrum of topics ranging from empowering themes, such as self-expression through drag, community support and positive body image, to challenges faced by the LGBTQ+ community, including homophobia, HIV and mental health. Employing an LLM allowed for nuanced exploration of these themes, overcoming the limitations of traditional word-based topic modeling.
Guardrails, not Guidance: Understanding Responses to LGBTQ+ Language in Large Language Models (Joshua Tint)
Language models have integrated themselves into many aspects of digital life, shaping everything from social media to translation. This paper investigates how large language models (LLMs) respond to LGBTQ+ slang and heteronormative language. Through two experiments, the study assesses the emotional content and the impact of queer slang on responses from models including GPT-3.5, GPT-4o, Llama2, Llama3, Gemma and Mistral. The findings reveal that heteronormative prompts can trigger safety mechanisms, leading to neutral or corrective responses, while LGBTQ+ slang elicits more negative emotions. These insights punctuate the need to provide equitable outcomes for minority slangs and argots, in addition to eliminating explicit bigotry from language models.
Dehumanization of LGBTQ+ Groups in Sexual Interactions with ChatGPT (Alexandria Leto, Juan Vásquez, Alexis Palmer, Maria Leonor Pacheco)
Given the widespread use of LLM-powered conversational agents such as ChatGPT, analyzing the ways people interact with them could provide valuable insights into human behavior. Prior work has shown that these agents are sometimes used in sexual contexts, such as to obtain advice, to role-play as sexual companions, or to generate erotica. While LGBTQ+ acceptance has increased in recent years, dehumanizing practices against minorities continue to prevail. In this paper, we hone in on this and perform an analysis of dehumanizing tendencies toward LGBTQ+ individuals by human users in their sexual interactions with ChatGPT. Through a series of experiments that model various concept vectors associated with distinct shades of dehumanization, we find evidence of the reproduction of harmful stereotypes. However, many user prompts lack indications of dehumanization, suggesting that the use of these agents is a complex and nuanced issue which warrants further investigation.
Leveraging Large Language Models in Detecting Anti-LGBTQIA+ User-generated Texts (Quoc-Toan Nguyen, Josh Nguyen, Van-Tuan Pham, William John Teahan)
Anti-LGBTQIA+ texts in user-generated content pose significant risks to online safety and inclusivity. This study investigates the capabilities and limitations of five widely adopted Large Language Models (LLMs)—DeepSeek-V3, GPT-4o, GPT-4o-mini, GPT-o1-mini, and Llama3.3-70B—in detecting such harmful content. Our findings reveal that while LLMs demonstrate potential in identifying offensive language, their effectiveness varies across models and metrics, with notable shortcomings in calibration. Furthermore, linguistic analysis exposes deeply embedded patterns of discrimination, reinforcing the urgency for improved detection mechanisms for this marginalised population. In summary, this study demonstrates the significant potential of LLMs for practical application in detecting anti-LGBTQIA+ user-generated texts and provides valuable insights from text analysis that can inform topic modelling. These findings contribute to developing safer digital platforms and enhancing protection for LGBTQIA+ individuals.
A Bayesian account of pronoun and neopronoun acquisition (Cassandra L Jacobs, Morgan Grobol)
A major challenge to equity among members of queer communities is the use of one's chosen forms of reference, such as personal names or pronouns. Speakers often dimiss errors in pronominal use as unintentional, and claim that their errors reflect many decades of fossilized mainstream language use, including attitudes or expectations about the relationship between one's appearance and acceptable forms of reference. Here, we propose a modeling framework that allows language use and speech communities to change over time, including the adoption of neopronouns and other forms for self-reference. We present a probabilistic graphical modeling approach to pronominal reference that is flexible in the face of change and experience while also moving beyond form-to-meaning mappings. The model critically also does not rely on lexical covariance structure to learn referring expressions. We show that such a model can account for individual differences in how quickly pronouns or names are integrated into symbolic knowledge and can empower computational systems to be both flexible and respectful of queer people with diverse gender
Non-archival
Some Myths About Bias: A Queer Studies Reading of Bias Evaluation and Mitigation Techniques in NLP (Filipa Calado)
This paper critically examines gender bias in large language models (LLMs) by integrating concepts from Queer Studies, particularly the theory of gender performativity and the critique of binary forms. It argues that many existing bias detection and mitigation techniques in Natural Language Processing (NLP), such as the Word Embedding Association Test (WEAT) and gender swapping methods, rely on outdated conceptualizations of gender, which take for granted the gender binary as a symmetrical and stable form. Drawing from Queer Studies, the paper highlights three "myths" about gender bias: that bias can be excised, that it is categorical, and that it can be leveled. Due to their operationalizing of the gender binary, each of these myths effectively reduce and flatten bias into a measure that fails to represent real-world workings of semantics, discrimination, and prejudice. The paper concludes by suggesting that bias mitigation in NLP should focus on amplifying diverse gender expressions and incorporating non-binary perspectives, rather than attempting to neutralize or equalize them. By reworking that which is outside the binary form, against which the binary defines itself, one may fashion more inclusive and intersectional approaches to mitigating bias in language systems.
Do language models practice what they preach? Examining language ideologies about gendered language reform encoded in LLMs (Julia Watson, Sophia Lee, Barend Beekhuizen, Suzanne Stevenson)
We study language ideologies in text produced by LLMs through a case study on English gendered language reform (related to role nouns like congressperson/-woman/-man, and singular they). First, we find political bias: when asked to use language that is “correct” or “natural”, LLMs use language most similarly to when asked to align with conservative (vs. progressive) values. This shows how LLMs’ metalinguistic preferences can implicitly communicate the language ideologies of a particular political group, even in seemingly non-political contexts. Second, we find LLMs exhibit internal inconsistency: LLMs use gender-neutral variants more often when more explicit metalinguistic context is provided. This shows how the language ideologies expressed in text produced by LLMs can vary, which may be unexpected to users. We discuss the broader implications of these findings for value alignment.
3rd Position, X (Chris Acorne)
This work tells the story of a young girl who dreams of becoming a ballerino and critiques the rigid gender binary emphasized in ballet. The colors behind the ballet shoes represent the nonbinary pride flag, sending a message of transcending the gender dichotomy. The title “3rd Position, X” refers to a pose often associated with male ballet dancers, and ‘X’ symbolizes a third gender. The painting explores the anguish of a young girl aspiring to become a ballerino, visually examining how the gender dichotomy affects one’s dreams and identity, while questioning the meaning of true inclusion.
Call for Contributions (CLOSED)
We are excited to announce our call for contributions to the Queer in AI Workshop at the 2025 NAACL Conference. We are accepting research papers, extended abstracts, position papers, opinion pieces, surveys, and artistic expressions on queer issues in NLP and Linguistics. We also welcome contributions about general topics in NLP and Linguistics authored by queer folks. We also highly encourage works highlighting New Mexican queer cultures and linguistic elements. Accepted contributions will be invited to present at the Queer in AI workshop during the 2025 NAACL Conference.
This workshop has two tracks: archival and non-archival. The archival versions of the workshops should adhere to ACL format and accepted works will be published in ACL anthology.
Archival submissions
Papers submitted to this track will be peer-reviewed and considered for submission in ACL anthology. The papers should be in English. Paper submissions must use the official ACL style templates, which are available here (Latex and Word). Please follow the general ACL paper formatting guidelines available here. You can either submit a short work of 4 pages or a long paper of 8 pages. We invite archival submissions in the following tracks:
Queer linguistics: This track invites submissions of studies related to the language around gender and sexuality. For example, the application of ideas from queer theory to language research, or providing an overview of the discursive formation of heteronormativity.
Queerness and NLP: This track invites submissions at the intersection between NLP and queerness. Possible themes could be the usage of NLP to analyze queer language or discovering flaws behind NLP methodologies which may harm queer folks.
Queer activism and D&I: This track invites submissions related to issues with queer inclusivity in NLP events and solutions to increase it. We also invite stories and strategies for queer activism in tech and academia, as well as artefacts and tools to promote queer advocacy.
If you want to submit your findings paper / un-accepted paper to our workshop, you can also commit the reviews to our workshop.
Non-archival submissions
For this format, you can submit your work in the form of art, poetry, music, microblogs, tiktoks, or videos. You need to upload a PDF containing a summary or abstract of your work and a link to your work. You can also submit your findings paper, a late-breaking paper or an abstract detailing your work. You can submit this paper in any language that you prefer.
Important Dates
All deadlines are Anywhere on Earth.
Archival submission deadline: January 30, 2025 February 20, 2025
Non-archival, ARR, Findings submission deadline: February 20, 2025
Decisions due: March 1, 2025
Camera-ready submissions due from accepted authors: March 10, 2025
If you need help with your submission in the form of mentoring or advice, you can get in touch with us at queer-in-nlp@googlegroups.com.
Submissions and Formatting Instructions
Submission is electronic, using the OpenReview platform. All papers must follow the ACL Author Guidelines. All submissions should be anonymized. Please refrain from including personally identifying information in your submission.
All authors with accepted work will have FULL control over how their names appear in public listings of accepted submissions.
Mentorship
If you are writing a paper for the first time and need some help with your work, we strongly suggest you contact us. If you are willing to help first time authors, please feel free to indicate us by emailing us.
Contact Us
Email: queer-in-nlp [at] googlegroups.com
Organizers
Ankush Gupta (he/they) Ankush is a final-year B.Tech student in Computer Science at IIIT-Delhi. He focuses his research on Human-Centered AI and edge computing. His research involves utilising deep learning and natural language processing to develop algorithms that improve human interaction and mitigate user biases to make AI systems more inclusive and effective.
Amanda Bertsch (she/her) Amanda is a queer PhD student at Carnegie Mellon University. Her research focuses on machine learning for text generation, particularly for long context modeling.
Arjun Subramonian (they/them) Arjun is a queer PhD candidate at UCLA. Their research focuses on the ethics and unfairness of machine learning.
Yanan Long (he/they) Yanan is an independent research scientist, most recently affiliated with University of Chicago. His research focuses on AI for science, tabular learning, machine translation and mechanistic interpretability.