[{"content":" Personal website of Delfi Sol Pandiani, investigating and breaking codified binaries.\n[ABOUT] [ART] [RESEARCH] [PUBLICATIONS] [TEACHING] // LATEST CHAPTER The Hallucinated Subjects Oxford Intersections, 2026 PAPER From vulnerable data subjects to vulnerabilizing data practices FAccT 2026 PAPER QueerGen: How LLMs reflect societal norms on gender and sexuality EACL Findings 2026 ","date":"18 August 2026","externalUrl":null,"permalink":"/","section":"Broken Binaries","summary":" Personal website of Delfi Sol Pandiani, investigating and breaking codified binaries.\n[ABOUT] [ART] [RESEARCH] [PUBLICATIONS] [TEACHING] // LATEST CHAPTER The Hallucinated Subjects Oxford Intersections, 2026 PAPER From vulnerable data subjects to vulnerabilizing data practices FAccT 2026 PAPER QueerGen: How LLMs reflect societal norms on gender and sexuality EACL Findings 2026 ","title":"Broken Binaries","type":"page"},{"content":" Courses # Master\u0026rsquo;s # r.M.A. Research Master\u0026rsquo;s in Media Studies, University of Amsterdam # Aesthetic Resistances in the Age of Capture [2026-2027] Hormones as Media(ting) Technologies Tutorial [2026-2027] M.A. Media Studies Cultural Data \u0026amp; AI, University of Amsterdam # Cultural Data Analysis [2024-2027] Embedded Research Projects [2024-2027] Data Project [2024-2027] M.Sc. Artificial Intelligence, University of Amsterdam # AI for Society [2024-2025] Bachelor\u0026rsquo;s # B.A. in Gloabl Arts, Culture, and Politics (GACP), University of Amsterdam # Public AI [2024-2027] B.A. in Media Studies (Media and Information), University of Amsterdam # Data-Driven Research \u0026amp; Digital Humanities Lab [2024-2025] Coding the Humanities: B.A. in Media \u0026amp; Information [2024-2025] B.Sc. in Information Science for Management, University of Bologna # Web Technologies [2021-2023] B.Sc. in Biological Sciences, University of Bologna # Computer Skills [2021-2023] Supervision # PhD Level Supervision # Institute of Logic, Language, and Computation, University of Amsterdam # Montanaro, C. (2028). Co-supervision with D. Beraldo and T. Blanke. N’Daiye, B. (2029). Co-supervision with T. Blanke and P. Helm. Master’s Level Supervision # M.A. Media Studies Cultural Data \u0026amp; AI, University of Amsterdam # Alonso Sanz, A. (2026). Ornamental Bodies: Reading Patterns of Racial and Cultural Recognition through Ornamental Assembly in Synthetic Imagery Majumdar, J. (2026). All Present, None Accounted For: Entity Recognition for Bridging Archival Access Gaps Hanh, L. (2026). “Not Yet”: Queer Sabotage, Adversarial Resistance and the Colonial Politics of the Latent Space Universal Adversarial Perturbations as Critical Technical Practice in Stable Diffusion 1.5 Elatarazova, Y. (2026). Prompting for Exploration: How User-Framed Epistemic Uncertainty Shapes Exploratory Affordances in Archival Chatbots Mazzola, F. (2026). The Unmarked Worker: Identity Markedness and Visual Normativity in Text-to-Image Portraiture of Labour Huang, J. (2026). Algorithmic Depoliticisation in Digital Heritage: Conversational AI and Visual Collection Interfaces as Sites of Epistemic Injustice Pos, R. (2026). Political Pixels: Investigating the Visual Discourse of /pol/ and its Role in Building an Anonymous Collective Identity During the 2024 U.S. Election Cycle. Vosmeijer, J. (2026). \u0026ldquo;Total passoid victory\u0026rdquo;: Homonormativity and Transnormativity on 4chan’s /lgbt/. Menehbi, D. (2025). The Face That Haunts the Machine: Exposing the Ghosts in AI Vision. Klein, E. (2025). Synthetic Shadows: A Systematic Analysis of Postcolonial, Intersectional, and Beautification Biases in AI-Generated Imagery. Voorzanger, L. (2025). Death of Man, Death of Punk: The Representation of ‘Punk’ in Synthetic Imagery via Text-To-Image AI. Chella Raghavendran, S. (2025). (Da)Lit Memes for Brahmin Teens: Exploring Casteism on 4Chan. Silingardi, S. (2025). Misogyny Repeated and Reposted: A Computational and Critical Analysis of How Toxic Memes Enact Gender Ideology through Multimodal Speech Acts. M.Sc. Artificial Intelligence, University of Amsterdam # Langerak, K. (2025). Using expert knowledge graphs to improve toxic meme detection and explanation. M.A. Digital Humanities and Digital Knowledge, University of Bologna # Veggi, M. (2023). [Co-supervision] MyTISSE: Interactive systems as enhancement tools for the sense of care in color perception. Bachelor’s Level Supervision # B.Sc. Computer Science, University of Amsterdam # Jheeta, J. (2025) Benchmarking toxic symbol detection in multimodal memes. B.Sc. Artificial Intelligence, University of Amsterdam # Caballer, L. (2024). Patterns of toxic symbology in internet memes: Leveraging unsupervised clustering methods for the identification of structural and semantic patterns. ","date":"18 August 2026","externalUrl":null,"permalink":"/teaching/","section":"Broken Binaries","summary":" Courses # Master’s # r.M.A. Research Master’s in Media Studies, University of Amsterdam # Aesthetic Resistances in the Age of Capture [2026-2027] Hormones as Media(ting) Technologies Tutorial [2026-2027] M.A. Media Studies Cultural Data \u0026 AI, University of Amsterdam # Cultural Data Analysis [2024-2027] Embedded Research Projects [2024-2027] Data Project [2024-2027] M.Sc. Artificial Intelligence, University of Amsterdam # AI for Society [2024-2025] Bachelor’s # B.A. in Gloabl Arts, Culture, and Politics (GACP), University of Amsterdam # Public AI [2024-2027] B.A. in Media Studies (Media and Information), University of Amsterdam # Data-Driven Research \u0026 Digital Humanities Lab [2024-2025] Coding the Humanities: B.A. in Media \u0026 Information [2024-2025] B.Sc. in Information Science for Management, University of Bologna # Web Technologies [2021-2023] B.Sc. in Biological Sciences, University of Bologna # Computer Skills [2021-2023] Supervision # PhD Level Supervision # Institute of Logic, Language, and Computation, University of Amsterdam # Montanaro, C. (2028). Co-supervision with D. Beraldo and T. Blanke. N’Daiye, B. (2029). Co-supervision with T. Blanke and P. Helm. Master’s Level Supervision # M.A. Media Studies Cultural Data \u0026 AI, University of Amsterdam # Alonso Sanz, A. (2026). Ornamental Bodies: Reading Patterns of Racial and Cultural Recognition through Ornamental Assembly in Synthetic Imagery Majumdar, J. (2026). All Present, None Accounted For: Entity Recognition for Bridging Archival Access Gaps Hanh, L. (2026). “Not Yet”: Queer Sabotage, Adversarial Resistance and the Colonial Politics of the Latent Space Universal Adversarial Perturbations as Critical Technical Practice in Stable Diffusion 1.5 Elatarazova, Y. (2026). Prompting for Exploration: How User-Framed Epistemic Uncertainty Shapes Exploratory Affordances in Archival Chatbots Mazzola, F. (2026). The Unmarked Worker: Identity Markedness and Visual Normativity in Text-to-Image Portraiture of Labour Huang, J. (2026). Algorithmic Depoliticisation in Digital Heritage: Conversational AI and Visual Collection Interfaces as Sites of Epistemic Injustice Pos, R. (2026). Political Pixels: Investigating the Visual Discourse of /pol/ and its Role in Building an Anonymous Collective Identity During the 2024 U.S. Election Cycle. Vosmeijer, J. (2026). “Total passoid victory”: Homonormativity and Transnormativity on 4chan’s /lgbt/. Menehbi, D. (2025). The Face That Haunts the Machine: Exposing the Ghosts in AI Vision. Klein, E. (2025). Synthetic Shadows: A Systematic Analysis of Postcolonial, Intersectional, and Beautification Biases in AI-Generated Imagery. Voorzanger, L. (2025). Death of Man, Death of Punk: The Representation of ‘Punk’ in Synthetic Imagery via Text-To-Image AI. Chella Raghavendran, S. (2025). (Da)Lit Memes for Brahmin Teens: Exploring Casteism on 4Chan. Silingardi, S. (2025). Misogyny Repeated and Reposted: A Computational and Critical Analysis of How Toxic Memes Enact Gender Ideology through Multimodal Speech Acts. M.Sc. Artificial Intelligence, University of Amsterdam # Langerak, K. (2025). Using expert knowledge graphs to improve toxic meme detection and explanation. M.A. Digital Humanities and Digital Knowledge, University of Bologna # Veggi, M. (2023). [Co-supervision] MyTISSE: Interactive systems as enhancement tools for the sense of care in color perception. Bachelor’s Level Supervision # B.Sc. Computer Science, University of Amsterdam # Jheeta, J. (2025) Benchmarking toxic symbol detection in multimodal memes. B.Sc. Artificial Intelligence, University of Amsterdam # Caballer, L. (2024). Patterns of toxic symbology in internet memes: Leveraging unsupervised clustering methods for the identification of structural and semantic patterns. ","title":"TEACHING","type":"page"},{"content":"GOOGLE SCHOLAR \u0026gt;\nSelected Conference Proceedings # Martinez Pandiani, D. S., Streefkerk, E., Naudts, L., \u0026amp; Helm, P. (2026). From vulnerable data subjects to vulnerabilizing data practices: Navigating the protection paradox in AI-based analyses of platformized lives. In Proceedings of the 2026 ACM Conference on Fairness, Accountability, and Transparency (FAccT ’26). Association for Computing Machinery, New York, NY, USA, 106–128. https://doi.org/10.1145/3805689.3806735\nSosto, M., Martinez Pandiani, D. S., \u0026amp; Hollink, L. (2026). QueerGen: How LLMs reflect societal norms on gender and sexuality in sentence completion task. In V. Demberg, K. Inui, \u0026amp; L. Marquez (Eds.), Findings of the Association for Computational Linguistics: EACL 2026 (pp. 4305–4326). Rabat, Morocco: Association for Computational Linguistics. https://doi.org/10.18653/v1/2026.findings-eacl.225\nMartinez Pandiani, D. S.. (2026). Tracing a Memetic Journey: From South American Death Flights to Free Helicopter Ride Memes. In: AoIR Selected Papers of Internet Research. https://doi.org/10.5210/spir.v2024i0.15240\nMartinez Pandiani, D. S., Tjong Kim Sang, E., \u0026amp; Ceolin, D. (2026). OnToxKG: An ontology-based knowledge graph of toxic symbols and their manifestations. In: Verma, H., Bozzon, A., Mauri, A., Yang, J. (eds) Web Engineering ICWE 2025. Lecture Notes in Computer Science, vol 15749.Springer, Cham. https://doi.org/10.1007/978-3-031-97207-2_9\nTjong Kim Sang, E., Martinez Pandiani, D. S., \u0026amp; Ceolin, D. (2025). Evaluating locally run large language models on toxic meme analysis. In: Verma, H., Bozzon, A., Mauri, A., Yang, J. (eds) Web Engineering ICWE 2025. Lecture Notes in Computer Science, vol 15749. Springer, Cham. https://doi.org/10.1007/978-3-031-97207-2_10\nMartinez Pandiani, D. S., Lazzari, N., \u0026amp; Presutti, V. (2024). Stitching gaps: Fusing situated perceptual knowledge with vision transformers for high-level image classification. In Proceedings of the 20th International Conference on Semantic Systems. IOS Press. https://doi.org/10.3233/SSW240008\nPescarin, S., \u0026amp; Martinez Pandiani, D. S. (2022). Factors in the cognitive-emotional impact of educational environmental narrative videogames. In: De Paolis, L.T., Arpaia, P., Sacco, M. (eds) Extended Reality. XR Salento 2022. Lecture Notes in Computer Science, vol 13446. Springer, Cham. https://doi.org/10.1007/978-3-031-15553-6_8\nMartinez Pandiani, D. S., \u0026amp; Presutti, V. (2022). Coded visions: Addressing cultural bias in image annotation systems with the descriptions and situations ontology design pattern. In Proceedings of the 6th International Conference of Graphs and Networks in the Humanities 2022: Technologies, Models, Analyses, and Visualizations.\nMartinez Pandiani, D. S., \u0026amp; Presutti, V. (2021). Automatic modeling of social concepts evoked by art images as multimodal frames. In Proceedings of the Workshops and Tutorials held at LDK 2021 co-located with the 3rd Language, Data and Knowledge Conference.\nSelected Journal Articles # Martinez Pandiani, D. S., Tjong Kim Sang, E. \u0026amp; Ceolin, D. (2025). ‘Toxic’ memes: A survey of computational perspectives on the detection and explanation of meme toxicities. Online Social Networks and Media, 47, 100317. https://doi.org/10.1016/j.osnem.2025.100317\nMartinez Pandiani, D. S. (2024). The wicked problem of naming the intangible: Abstract concepts, binary thinking, and computer vision labels. Future Humanities, 2, e11. Wiley Online Library. https://doi.org/10.1002/fhu2.11\nCiroku, F., De Giorgis, S., Gangemi, A., Martinez-Pandiani, D. S., \u0026amp; Presutti, V. (2024). Automated multimodal sensemaking: Ontology-based integration of linguistic frames and visual data. Computers in Human Behavior, 150, 107997. https://doi.org/10.1016/j.chb.2023.107997\nMartinez Pandiani, D. S., Lazzari, N., van Erp, M., \u0026amp; Presutti, V. (2023). Hypericons for interpretability: decoding abstract concepts in visual data. International Journal of Digital Humanities, 5, 451–490 (2023). https://doi.org/10.1007/s42803-023-00077-8\nDaga, E., Asprino, L., Damiano, R., Daquino, M., Diaz Agudo, B., Gangemi, A., Kuflik, T., Lieto, A., Maguire, M., Marras, A. M., \u0026amp; Martinez Pandiani, D. S. (2022). Integrating citizen experiences in cultural heritage archives: Requirements, state of the art, and challenges. ACM Journal on Computing and Cultural Heritage (JOCCH), 15, 1–35. https://dl.acm.org/doi/full/10.1145/3477599\nBook Chapters # Brooke, S. J. M., \u0026amp; Martinez Pandiani, D. S. (2026). The hallucinated subjects: Gender, coherence, and the ontological politics of generative AI. In M. Detloff \u0026amp; T. J. Billard (Eds.), Oxford Intersections: Gender Justice. Oxford University Press. https://doi.org/10.1093/9780198960256.003.0045\nVeggi, M., Bonanno, V., Cerato, I., Ciortan, I.-M., Clay, A., Fiorenza, G., Martinez Pandiani, D. S., \u0026amp; Pescarin, S. (2026). Sense of care in the design of digital heritage applications. In S. Pescarin, C. Barandoni, A. Clay, G. Papadopoulos, \u0026amp; I. C. A. Sandu (Eds.), Chromatic Visions: Exploring Colour in Art, Archaeology and Digital Realities, Part I (pp. 128–161). Springer Nature Switzerland. https://doi.org/10.1007/978-3-032-07792-9_4\nMartinez Pandiani, D. S. (2024). Bridging the gap: Decoding abstract concepts in cultural heritage images. In F. Moral-Andrés, E. Merino-Gómez, \u0026amp; P. Reviriego (Eds.), Decoding Cultural Heritage: A Critical Dissection and Taxonomy of Human Creativity through Digital Tools (pp. 157–189). Springer, Cham. https://doi.org/10.1007/978-3-031-57675-1_8\nPanels # Messi, J.S.; Martinez Pandiani, D. S.; Streefkerk, E.; Sosto, M.; Hanna, L. (2026). Against Technosolutionism: Governing Platforms as Systems of Care. Panelist at SPUI25, University of Amsterdam, February 23, 2026. https://spui25.nl/programma/against-technosolutionism-governing-platforms-as-systems-of-care\nMartinez Pandiani, D. S. [Organizer]; Voto, C., Beretta, E. (2025). Creative frictions: Control, generative failure and speculative archives in GenAI. Panelist at GenAI \u0026amp; Creative Practices: Past, Present, and Future, University of Amsterdam, December 17–18, 2025.\nVoto, C.; Beretta, E.; Martinez Pandiani, D. S., Da Hora, N.; Lee-Morrison, L. (2025). AI enunciations: Redefining identity, space, and meaning across disciplines. Panelist at the Ethics and Aesthetics of Artificial Images Conference, Venice, May 8–10, 2025.\nvan der Haak, B.; van Noord, N.; Martinez Pandiani, D. S.; Ayub, A. (2025). Human-aligned AI in video. Panelist at University of Amsterdam Data Science Center (DSC) Away Day with the HAVA (Human-Aligned Video AI Lab) at the Eye Film Institute, Amsterdam, June 18, 2025.\nSelected Lectures and Talks # Helm, P. \u0026amp; Martinez Pandiani, D. S. (2026). Rendering Vulnerability: AI, Trauma, and the Infrastructures of Recognition. Keynote lecture at International Conference on Rethinking Trauma, Babeș-Bolyai University, Cluj-Napoca, Romania, September 2-4, 2026. https://rethinkingtrauma.conference.ubbcluj.ro/keynotes/\nMartinez Pandiani, D. S. (2026). Algorithmic Mediation of Intimacy as/and Vulnerability. Guest lecture at New Media and the Digitalisation of Everyday Life Summer School, University of Amsterdam, The Netherlands, August 6, 2026.\nMartinez Pandiani, D. S. (2026). Cultural boundaries in latent space: Iterative image prompting as method and meaning-making. STS NL 2026 (Deep Learning and Culture), Twente, The Netherlands, April 15-17, 2026.\nMartinez Pandiani, D. S. (2026). Memes, memory, and digital coloniality: Tracing the global afterlives of South American death flights. Global Digital Humanities Symposium (GDHS) 2026, Virtual and In-Person, Michigan, United States, 13-17 April 2026.\nMartinez Pandiani, D. S. (2025). Follow the glitch: Generative failures and creative navigation. GenAI \u0026amp; Creative Practices: Past, Present, and Future, University of Amsterdam, The Netherlands, December 2025.\nMartinez Pandiani, D. S. (2025). Tracing a memetic journey: From South American death flights to free helicopter ride memes. Association of Internet Researchers (AoIR 2025), Niteroi, Brazil, 15-18 October 2025.\nMartinez Pandiani, D. S. (2025). AI, culture, and critical thinking: Navigating opportunities, biases, and ethics in the digital age. AMI AGM: Vision, Values and Voices — Annual General Meeting of the Association Montessori Internationale, Delft, The Netherlands, 12 April 2025.\nMartinez Pandiani, D. S. (2025). Enacting vulnerability: AI, childhood, and the co-construction of vulnerable data subjects. Institute for Advanced Study (IAS), Amsterdam, 21 May 2025. Kick-off lecture of the IAS-DSC fellowship.\nMartinez Pandiani, D. S. (2024). Leveraging LLMs for detection and interpretation of toxic memes. Internet Research with Foundation Models Workshop at CAT4SMR (Capture and Analysis Tools for Social Media Research), Amsterdam, 3 September 2024.\nMartinez Pandiani, D. S. (2024). Relación texto-imagen en memes: Construcción (en curso) de un modelo de IA para la detección automática de “narrativas tóxicas” [Text-image relationship in memes: (Ongoing) Construction of an AI model for automatic detection of \u0026ldquo;toxic narratives\u0026rdquo;]. V Congreso Internacional PRISMA: Lenguaje e Inteligencia Artificial: Aproximaciones lingüísticas y desafíos interdisciplinarios, Universidad Adolfo Ibáñez, Santiago, Chile, 28-30 May 2024.\nMartinez Pandiani, D. S. (2024). Toxic memes workshop. AI, Media \u0026amp; Democracy Lab, University of Amsterdam, Amsterdam, 2 April 2024.\nMartinez Pandiani, D. S. (2024). Coding the encoder: Situating subjective and contextual aspects in high-level image annotations. International Conference on Reimagining Annotation for Multimodal Cultural Heritage, Rennes 2 University and the MSHB (Maison des Sciences de l\u0026rsquo;Homme en Bretagne), France, 7-9 Feb 2024.\nMartinez Pandiani, D. S. (2023). Collapsing meaning in hypericon images: Abstract concepts, distributed reality, and perceptual bias. Operational Imaginaries: Images of Abstraction, Operational Media and Experience, Amsterdam School for Cultural Analysis, University of Amsterdam, Netherlands, 25-26 May 2023.\nMartinez Pandiani, D. S. (2023). The wicked problem of naming the intangible: Abstract concepts, binary thinking, and computer vision labels. FACEing Binarism. Towards a more equitable AI symposium, Vrije Universiteit Amsterdam, Netherlands, 30 March 2023.\nMartinez Pandiani, D. S. (2022). Seeing the intangible: Interpretable visual understanding of high-level abstract concepts. User-Centric Data Science, Vrije Universiteit Amsterdam, Netherlands, 25 November 2022.\nMartinez Pandiani, D. S. (2022). Tip of the iceberg: Abstract concepts in high-level visual understanding. IMT School for Advanced Studies\u0026rsquo; seminar: Semantic Representation, Abstractness and Abstraction, Lucca, Italy, 27 April 2022.\nMartinez Pandiani, D. S. (2022). Coded visions: Cultural bias in assigning meaning to digital images. Digital Humanities \u0026amp; Digital Communication Spring Seminars, Università di Modena e Reggio Emilia, Modena, Italy, 28 March 2022.\n","date":"1 January 2024","externalUrl":null,"permalink":"/publications/","section":"Broken Binaries","summary":"GOOGLE SCHOLAR \u003e\nSelected Conference Proceedings # Martinez Pandiani, D. S., Streefkerk, E., Naudts, L., \u0026 Helm, P. (2026). From vulnerable data subjects to vulnerabilizing data practices: Navigating the protection paradox in AI-based analyses of platformized lives. In Proceedings of the 2026 ACM Conference on Fairness, Accountability, and Transparency (FAccT ’26). Association for Computing Machinery, New York, NY, USA, 106–128. https://doi.org/10.1145/3805689.3806735\n","title":"PUBLICATIONS","type":"page"},{"content":"The Methodological Framework: Following the GlitchPipeline StageComputational AssumptionThe Manifested \u0026ldquo;Glitch\u0026quot;Critical \u0026amp; Tactical ResponseOperationalizationHuman concepts can be mapped cleanly into discrete, computable taxonomies.Complex subjectivities break down; binary grids misclassify and erase non-normative experience.Exposing Epistemic Violence: Tracing how classification systems inherit colonial and heterocisnormative assumptions.Latent ModelingHigh-dimensional embeddings neutrally represent semantic similarity.Latent geometries encode historical power asymmetries, stereotyping, and docility regimes.Latent Space Deconstruction: Auditing the unprompted aesthetic and semantic biases of generative models.Inference \u0026amp; AlignmentAlignment algorithms produce safe, objective, and neutral conversational personas.Models hallucinate rigid human and machinic subjects to enforce coercive gendered coherence.Queer Refusal \u0026amp; Ungovernability: Strategically breaking model predictability via auto-experimentation and opacity.Intervention / AI4SGApplying AI to detect harm inherently protects vulnerable populations.The Protection Paradox: Interventions amplify exposure, surveillance, and narrative fixing.Reflexive Practice: Shifting focus from \u0026ldquo;vulnerable subjects\u0026rdquo; to the ethics of \u0026ldquo;vulnerabilizing practices.\u0026rdquo;\nMethodological Stance: \u0026ldquo;Following the Glitch\u0026rdquo; # I am especially interested in how these operationalizations are used to enact and echo power dynamics, including gendered and colonial dynamics, and how they get embedded in systems of governance (like content moderation) and increasingly surveillance.\nComputational Assumption Manifested \u0026ldquo;Glitch\u0026rdquo; Critical \u0026amp; Tactical Response Human concepts map cleanly into discrete, computable taxonomies. Complex subjectivities break down; binary grids misclassify non-normative experience. Exposing Epistemic Violence: Tracing how classification systems inherit colonial and heterocisnormative assumptions. High-dimensional embeddings neutrally represent semantic proximity. Latent geometries encode historical power asymmetries and docility regimes. Latent Space Deconstruction: Auditing the unprompted aesthetic and semantic biases of generative models. Alignment algorithms produce safe, objective, and neutral conversational personas. Models hallucinate rigid human and machinic subjects to enforce coercive gendered coherence. Queer Refusal \u0026amp; Ungovernability: Strategically breaking model predictability via opacity (Glissant) and fugitivity (Bey). Applying AI to detect harm inherently protects vulnerable populations. The Protection Paradox: Interventions amplify exposure, surveillance, and narrative fixing. Reflexive Practice: Shifting focus from \u0026ldquo;vulnerable subjects\u0026rdquo; to the ethics of \u0026ldquo;vulnerabilizing practices.\u0026rdquo; ","externalUrl":null,"permalink":"/glitch/","section":"Broken Binaries","summary":"The Methodological Framework: Following the GlitchPipeline StageComputational AssumptionThe Manifested “Glitch\"Critical \u0026 Tactical ResponseOperationalizationHuman concepts can be mapped cleanly into discrete, computable taxonomies.Complex subjectivities break down; binary grids misclassify and erase non-normative experience.Exposing Epistemic Violence: Tracing how classification systems inherit colonial and heterocisnormative assumptions.Latent ModelingHigh-dimensional embeddings neutrally represent semantic similarity.Latent geometries encode historical power asymmetries, stereotyping, and docility regimes.Latent Space Deconstruction: Auditing the unprompted aesthetic and semantic biases of generative models.Inference \u0026 AlignmentAlignment algorithms produce safe, objective, and neutral conversational personas.Models hallucinate rigid human and machinic subjects to enforce coercive gendered coherence.Queer Refusal \u0026 Ungovernability: Strategically breaking model predictability via auto-experimentation and opacity.Intervention / AI4SGApplying AI to detect harm inherently protects vulnerable populations.The Protection Paradox: Interventions amplify exposure, surveillance, and narrative fixing.Reflexive Practice: Shifting focus from “vulnerable subjects” to the ethics of “vulnerabilizing practices.”\n","title":"","type":"page"},{"content":"Abstract Labels and the AI Hype in Computer Vision\nWhat kinds of subjective, cultural labels are we asking AI models to assign to visual culture? How well do these work, and what biased visions of the world do they reproduce?\nSituated Ground Truths and Plural Perspectives\nWhat are we teaching AI as \u0026ldquo;truth\u0026rdquo; to emulate? How do we technically operationalize Haraway\u0026rsquo;s idea of \u0026ldquo;situated knowledge\u0026rdquo;\u0026ndash;partial, situated perspectives\u0026ndash;for AI?\nToxic Memes and Content Moderation\nMemes and other viral cultural data can spread dangerous messages. But exactly what do we mean with online toxicity, and how are morality and normativity operationalized?\nDepictions of the AI/Human Dynamic\nAre we the chimpanzees of the AI era? No, but contemporary media is using evolutionary rhetoric to depict AIs as an \u0026ldquo;evolution\u0026rdquo; of human beings. Why and how?\nLATENT IDENTITIES # Identity Labels in Latent Spaces: Generative AI and Stereotypes\nInvestigating how Generative AI models learn and represent identity labels in their latent spaces, with a focus on detecting and mitigating biases and stereotypes.\nAnalyzing the types of prompts, concepts, and labels people use to describe their identities and comparing them with labels assigned by AI.\nLatent Identities is a research direction that seeks to understand how Generative AI models learn and represent identity labels in their latent spaces. By investigating the ways individuals self-identify and the prompts and labels they use to describe their identities, researchers can uncover the elements that people consider vital but are often overlooked in AI training datasets. This information can be used to develop more inclusive and accurate representations of diverse identities, mitigating the biases and stereotypes that exist in current AI models.\n​\nThe goal is to identify and challenge the stereotypical biases that are embedded in Generative AI models. By comparing self-definitions with labels assigned by AI, researchers can highlight the biases and inaccuracies that exist in these models. This comparative analysis will help researchers understand what AI gets \u0026ldquo;right,\u0026rdquo; what it gets \u0026ldquo;wrong,\u0026rdquo; and what it fails to consider, enabling them to develop more fair and inclusive AI systems.\n​\nOne of the key aspects of this research is its focus on the representation of underrepresented groups. By examining the ways in which AI models learn to represent these groups, researchers can identify and challenge the visual stereotypes that are perpetuated through media usage and stock photo representations. This research can inform the development of alternative image banks and tools that are generated with citizen involvement, leading to more inclusive representations in mainstream media. Ultimately, I aim to create AI systems that promote diversity and inclusion, rather than marginalizing underrepresented communities.\nTOXICITY # Operationalization of Toxicity and Literacy: Toxic Symbology in Memes\nDeveloping AI models for detecting and analyzing toxicity in online platforms, with a focus on identifying patterns of toxic symbology in memes.\nAdvocating for digital literacy skills, particularly in relation to online platforms and social media, to promote a more positive and respectful online culture.\nThis research direction seeks to understand the complex and dynamic phenomenon of toxicity online. This research technically and theoretically explores the operationalization of concepts like toxicity, hatefulness, harmfulness, ethics, moderation, and extremism. By combining data-driven approaches with media studies insights, I aim to develop a deeper understanding of the ways in which AI models learn to recognize and mitigate toxic content.\n​\nOne key area of focus for this research is the development of software that can detect and explain toxic symbology in online platforms. This includes not only identifying toxic content, but also providing insights into the underlying reasons why it is toxic. By examining cases of memes and toxic symbology, such as the extremist, racist, and other hateful memes that spread on platforms like 4chan and Reddit, we can gain a better understanding of how toxicity is perpetuated and how it can be mitigated. This research also involves experimental computer science work, including the development of Retrieval Augmented Generation (RAG) based systems for toxicity detection and explanation.\n​\nThis research direction is also focused on the development of digital literacy skills, particularly in relation to online platforms and social media. By studying patterns of toxic symbology in memes, I aim to identify common themes and motifs that contribute to online virality. This information can be used to develop educational programs and tools that help users recognize and resist toxic content, promoting a more positive and respectful online culture.\nVULNERABILITY # Empirical Ethics of Vulnerable Data Subjects: Using AI on Childfluencers\nUsing AI to analyze family vlog content, identifying patterns of potential exploitation to inform policymakers and the public about children as monetized data subjects.​\nBalancing public interest with individual rights, ensuring that the use of AI is ethically sound and respectful of vulnerable data subjects, such as children.\nThis research direction is focused on the empirical ethics of data science when it comes to vulnerable data subjects, with a specific case study on the use of AI to analyze family vlog content for online child protection. This research aims to identify empirical patterns of potential exploitation on child influencers, using AI and other computational data science methods. The goal is to provide quantifiable evidence of overexposure and exploitation, highlighting potential harms and risks to children\u0026rsquo;s online presence.\n​\nBy analyzing family vlog content, researchers can identify patterns of overexposure, including to harmful audiences. This information can be used to inform policymakers, advocacy groups, and the public about the normalization of children as monetized data subjects, prompting critical discussions on ethical dilemmas and potential harms. The research also aims to drive policy changes by supporting legal frameworks, such as the Illinois law requiring parents to set aside a portion of earnings for child influencers.\n​\nHowever, this research direction also raises ethical concerns, particularly with regards to the use of AI on vulnerable data subjects, including children. Part of this research direction is the development of empirical ethics frameworks to consider potential harms, such as a lack of explicit consent, emotional or social impact, normalization of surveillance, legal or platform-specific repercussions, privacy invasion, and misuse of findings.\n","externalUrl":null,"permalink":"/research_humandigitalist/","section":"Broken Binaries","summary":"Abstract Labels and the AI Hype in Computer Vision\nWhat kinds of subjective, cultural labels are we asking AI models to assign to visual culture? How well do these work, and what biased visions of the world do they reproduce?\n","title":"","type":"page"},{"content":" ABOUT # \u0026gt; COLLECTING DATA FROM WEB.... [OK] \u0026gt; FLATTENING CONTRADICTION AND COMPLEXITY... [OK] \u0026gt; CRAFTING A SUFFICIENTLY COHERENT STORY... [OK] \u0026gt; COHERENT STORY READY_ DELFI SOL PANDIANI # Delfi (Sol Martinez) Pandiani [they//them] is an academic and creative researcher studying, playing with, and breaking codified binaries. Their research explores how abstract cultural concepts — especially identity, toxicity, and vulnerability — are computationally modeled in datafied environments. They audit, develop, and critique AI systems, aiming to surface new ways of thinking about power, representation, and technological change. They also maintain an artistic practice, especially painting, mixed media, and creative writing.\nDelfi is Assistant Professor of Cultural Data Analysis at the University of Amsterdam (UvA), appointed across the Department of Media Studies and the Institute for Logic, Language, and Computation (ILLC). They co-coordinate the Cultural Data \u0026amp; AI (CDAI) track of the Master\u0026rsquo;s in Media Studies at the UvA, which combines critical inquiry with data science.\nThey are co-founder of the Queer and Feminist Informatics Network (QFIN), and have been developing creative workshops that combine computational experimentation and critical theory with artistic practice — such as adversarial drag against facial and gender surveillance, and other interventions that play with, challenge, and attempt to break codified binaries (male/female, public/private, toxic/safe, white/colored, nature/culture).\nPreviously, Delfi was a post-doctoral researcher at the Human-Centered Data-Analytics (HCDA) group at Centrum Wiskunde \u0026amp; Informatica (CWI) in Amsterdam. They have a Ph.D. in Computer Science and Engineering (DISI) and a M.A. in Digital Humanities and Digital Knowledge (DHDK) from the Università di Bologna (Italy), and a Bachelor\u0026rsquo;s degree in Human Evolutionary Biology (HEB) (summa cum laude) with a minor in Gender and Sexuality Studies from Harvard University (USA).\nGITHUB \u0026gt;\nGOOGLE SCHOLAR \u0026gt;\nLINKEDIN \u0026gt;\nEDUCATION # \u0026gt; ACCESSING EDUCATION DATABASES... [OK] \u0026gt; TRADITIONAL EDUCATION RECORDS IDENTIFIED... [OK] \u0026gt; COHERENT ACADEMIC PATH READY_ Period Institution Degree Thesis Focus areas 2020–2024 Università di Bologna Ph.D. Computer Science \u0026amp; Engineering Mind the Gap(s): Cognitive-Inspired AI for High-Level Visual Sensemaking. Towards Abstract Concept Image Classification Explainable AI, Knowledge Engineering, Computer Vision, Digital Humanities 2018–2020 Università di Bologna M.A. Digital Humanities \u0026amp; Digital Knowledge Semantic \u0026amp; Interactive Technologies for Civic Education: Ancient Classical Polychromy as a Case Study Semantic Technologies, AR, Civic Education, Chromophobia, White Supremacy 2013–2017 Harvard University B.A. Human Evolutionary Biology (summa cum laude) Reconciliation in Homo sapiens: Behavioral Perspectives on the Human Post-Conflict Period Human Behavioral Coding, Data Analysis, Pro-sociality, Primate Behavior ACADEMIC BIO (150 words) # Delfi Sol (Martinez) Pandiani [they//them] is Assistant Professor of Cultural Data Analysis at the University of Amsterdam (UvA), at the Institute for Logic, Language and Computation (ILLC) and the Department of Media Studies. Their work explores how abstract social concepts—such as identity, toxicity, and vulnerability—are negotiated and computationally modeled in datafied environments. Drawing on critical AI studies, computer vision, digital humanities, and media and queer theory, they specialize in auditing, developing, and critiquing AI systems. They challenge binary assumptions like public/private and nature/culture, aiming to surface new ways of thinking about power, representation, and technological change. At the UvA, Delfina co-coordinates the MA track Cultural Data \u0026amp; AI, which combines critical inquiry with data science, and co-founded the Queer and Feminist Informatics Network (QFin). They hold a PhD in Computer Science, an MA in Digital Humanities from the University of Bologna, and a BA in Human Evolutionary Biology from Harvard University.\n","externalUrl":null,"permalink":"/about/","section":"Broken Binaries","summary":"ABOUT # \u003e COLLECTING DATA FROM WEB.... [OK] \u003e FLATTENING CONTRADICTION AND COMPLEXITY... [OK] \u003e CRAFTING A SUFFICIENTLY COHERENT STORY... [OK] \u003e COHERENT STORY READY_ DELFI SOL PANDIANI # Delfi (Sol Martinez) Pandiani [they//them] is an academic and creative researcher studying, playing with, and breaking codified binaries. Their research explores how abstract cultural concepts — especially identity, toxicity, and vulnerability — are computationally modeled in datafied environments. They audit, develop, and critique AI systems, aiming to surface new ways of thinking about power, representation, and technological change. They also maintain an artistic practice, especially painting, mixed media, and creative writing.\n","title":"ABOUT","type":"page"},{"content":"","externalUrl":null,"permalink":"/art/","section":"Art","summary":"","title":"Art","type":"art"},{"content":"","externalUrl":null,"permalink":"/categories/","section":"Categories","summary":"","title":"Categories","type":"categories"},{"content":" navel of the world (2018) # DOWNLOAD\nthe capsule (2017) # DOWNLOAD\n","externalUrl":null,"permalink":"/art/creative-writing/","section":"Art","summary":" navel of the world (2018) # DOWNLOAD\nthe capsule (2017) # DOWNLOAD\n","title":"CREATIVE WRITING","type":"art"},{"content":" Where Do We Go From Here # Harvard University, 2017 # Where Do We Go From Here — a structure of 14 three-sided columns made of Plexiglas and aluminum — was a public art installation displayed on Harvard University\u0026rsquo;s campus. It was conceived in the wake of the results of the Association of American Universities-led sexual conduct survey (\u0026ldquo;Campus Climate Survey on Sexual Assault and Sexual Misconduct\u0026rdquo;), which was distributed to every degree-seeking student at Harvard during the spring of 2015. Along with co-creator Devon Guinn 17\u0026rsquo;, we conceived of this interactive artwork to engage the University in exploring the devastating statistics on sexual assault that the report presented. Many students participated in fabricating and/or interacting with these columns, which were hosted previously by the Harvard residential Houses and student common spaces for a two-week period. Students were invited to write and draw on the columns, or express private thoughts on notes placed inside them. The completed columns were later presented, for viewing only, as a single art installation for a week during the Harvard ARTS FIRST Festival. They were placed in Harvard\u0026rsquo;s iconic Tercentenary Theater. The project was supported financially by the Harvard Office for the Arts and logistically by OFA Director Jack Megan, along with the assistance of artist Ross Miller \u0026lsquo;77 as technical adviser and consultant.\nMore information about the project:\nHarvard Gazette: \u0026ldquo;Making art, making community: Students organize an art installation in response to the sexual assault survey\u0026rdquo; (April 13, 2016). https://news.harvard.edu/gazette/story/2016/04/making-art-making-community/ Harvard Crimson: \u0026ldquo;Students\u0026rsquo; Art Installation Asks Harvard to Reflect on Sexual Assault\u0026rdquo; (April 13, 2016). https://www.thecrimson.com/article/2016/4/13/art-installation-on-sexual-assault/ [Phase] Transitioning # BBP Gallerie, \u0026ldquo;Here, without -\u0026rdquo;, 2015 # Resident artist to BBP Gallerie in its yearlong \u0026ldquo;Here, without -\u0026rdquo; curation project, culminating in an exhibition at Harvard\u0026rsquo;s Carpenter Center for the Visual Arts. Boston-based artists were matched to Palestinian and Israeli artists. I developed an individual, mixed-media installation, [Phase] Transitioning, composed of wax sculptures and a video projection. The piece examines the experience of desensitization, both in artistic processes and in the experiences of war and genocide.\n","externalUrl":null,"permalink":"/art/installations/","section":"Art","summary":" Where Do We Go From Here # Harvard University, 2017 # Where Do We Go From Here — a structure of 14 three-sided columns made of Plexiglas and aluminum — was a public art installation displayed on Harvard University’s campus. It was conceived in the wake of the results of the Association of American Universities-led sexual conduct survey (“Campus Climate Survey on Sexual Assault and Sexual Misconduct”), which was distributed to every degree-seeking student at Harvard during the spring of 2015. Along with co-creator Devon Guinn 17’, we conceived of this interactive artwork to engage the University in exploring the devastating statistics on sexual assault that the report presented. Many students participated in fabricating and/or interacting with these columns, which were hosted previously by the Harvard residential Houses and student common spaces for a two-week period. Students were invited to write and draw on the columns, or express private thoughts on notes placed inside them. The completed columns were later presented, for viewing only, as a single art installation for a week during the Harvard ARTS FIRST Festival. They were placed in Harvard’s iconic Tercentenary Theater. The project was supported financially by the Harvard Office for the Arts and logistically by OFA Director Jack Megan, along with the assistance of artist Ross Miller ‘77 as technical adviser and consultant.\n","title":"INSTALLATIONS","type":"art"},{"content":" Relieves Inseguros (2016) # Plaster, acrylic, and paper on wooden trunks. Each trunk between 30 and 50 cm in diameter.\nOmbligo Caleidoscópico (2018) # Chromatic photographic collage, all images captured in Rapa Nui (Easter Island) between 2017-2018.\n","externalUrl":null,"permalink":"/art/mixed-media/","section":"Art","summary":" Relieves Inseguros (2016) # Plaster, acrylic, and paper on wooden trunks. Each trunk between 30 and 50 cm in diameter.\n","title":"MIXED MEDIA","type":"art"},{"content":" Ayúda-me # Harvard College Women\u0026rsquo;s Center, 2014 # This 7 x 7 ft (2,13 x 2,13 m) mural was painted to honor the experiences of Latina immigrant victims of domestic violence in the U.S. It was conceived, designed and painted on December 2014 in the basement of Lowell House, Harvard University. It was later exhibited in the Conference Room of the Harvard College Women\u0026rsquo;s Center. The mural depicts the obstacles that women face when seeking help, which include cultural, structural, situational, and institutional barriers. Latina immigrant victims find themselves in an especially precarious position because of intersecting axes of power that frame their experiences: gender, race, ethnicity, national origin, immigration status and class, among others.\nAn intersectional approach to the reproductive justice issue of domestic abuse and violence is necessary for effective prevention, intervention and advocacy programs. Understanding the multiplicative effects of different social standings and the precarious place that Latina immigrants populate when facing Intimate Partner Violence (IPV) needs an approach that considers all levels of barriers: from the micro to the macro, from personal convictions to institutional policies.\nStarring In # Harvard Sustainability Office, 2017 # This 9 x 6 ft (2,75 x 1,82 m) mural was painted for the Harvard Sustainability Office as a reminder ti its employees of the connection between elite institutions, capitalism, narcissism and climate change. The following text accompanies the piece:\nIn working to save our planet, some argue we should learn from ants. Ants have a perfect society, they say, as they all work—and do their small part to their 100%.\nIt is unfortunate that when human youngsters discover the power of the magnifying glass, some deliberately incinerate those dedicated workers.\nIt is even more unfortunate that when human grownups discover that power, some carelessly incinerate their own.\nIn the blindness of our collective narcissism, we are roasting ourselves. Too intoxicated with our own reflection, we are failing to notice that which, on its way far away from us, offers genuine splendor.\n[A respected analyst at a major bank recently coined the term vanity capital to refer to the crucial market of conspicuous consumption, and—-have no fear!-—it is an exponentially growing one.]\nSee the official site of the mural in the Harvard Sustainability Office website.\nNútre-me # The STEP Centre, Kingston, Jamaica, 2015 # This 27 x 9 ft (8,23 x 2,75 m) mural was painted in the summer of 2015 in Kingston to honor the teachers and other individuals who nourish the minds of the young. It was conceived, designed and painted pro bono in June-July 2015 at The STEP Centre in the heart of Kingston, Jamaica. The STEP Centre is a school for the therapy, education, and parenting of children with multiple special needs (http://www.thestepcentre.com/). The mural was designed as a background to and in communication with the school\u0026rsquo;s vibrant life.\nRongo Rongo # Hare Nga Poki, Rapa Nui (Easter Island), 2018 # This mural encompasses all the walls of a children\u0026rsquo;s library in Rapa Nui (Easter Island). The library was painted pro bono, in collaboration and with inspiration from the local language, history, culture, and children\u0026rsquo;s stories. Rongo Rongo is Rapa Nui\u0026rsquo;s uncracked system of writing. Numerous attempts at decipherment have been made, yet it is still not clear what the tablets found on the island read. Nevertheless, Rongo Rongo is a holder of ancestral knowledge and evokes the importance of maintaining the Rapa Nui language and culture alive, including by teaching young children how to speak, write and read in Rapa Nui. For this reason, the children\u0026rsquo;s library\u0026rsquo;s door features a moai-held rongo rongo tablet. The time in the island and provision of supplies was supported by a Gardner Peabody Fellowship entrusted to Delfi 2017-2018.\n","externalUrl":null,"permalink":"/art/murals/","section":"Art","summary":" Ayúda-me # Harvard College Women’s Center, 2014 # This 7 x 7 ft (2,13 x 2,13 m) mural was painted to honor the experiences of Latina immigrant victims of domestic violence in the U.S. It was conceived, designed and painted on December 2014 in the basement of Lowell House, Harvard University. It was later exhibited in the Conference Room of the Harvard College Women’s Center. The mural depicts the obstacles that women face when seeking help, which include cultural, structural, situational, and institutional barriers. Latina immigrant victims find themselves in an especially precarious position because of intersecting axes of power that frame their experiences: gender, race, ethnicity, national origin, immigration status and class, among others.\n","title":"MURALS","type":"art"},{"content":" San Giorgio Blue [Crack Haha] (2022) — Acrylic on canvas, 40 × 30 cm., with Elisa Santos. (Non) Ho Pelli Sulla (2019) — Acrylic on canvas, 70 × 50 cm. Dalí (2017) — Acrylic on canvas, 100 × 120 cm. Ciggie Break (2021) — Acrylic on canvas, 30 × 30 cm. Chico Blue (2026) — Acrylic on canvas, 40 × 50 cm. Bearing Body Series (2017), oil on canvas. Bearing Body Series (2017) # Four-part series that isolates fragments of the bearing (the feminized as potentially pregnant) figure caught mid-gesture, attending to the body as a site of weight, posture, and the quiet labor of holding a/one self together.\nI: The Bearer (2017) — Oil on canvas, 100 × 70 cm. II: Ideal Bearing (2017) — Oil on canvas, 70 × 100 cm. III: Actual Bearing (2017) — Oil on canvas, 100 × 100 cm. IV: The Beared [From the Belly Button] (2017) — Oil on canvas, 70 × 70 cm. A Flor de Piel series. A Flor de Piel Series (2017) # A series of oil canvases attending to skin, surface, and the politics of the body\u0026rsquo;s edges.\nPedicure (2017) — Oil on canvas, 50 × 50 cm. Plena (2017) — Oil on canvas, 80 × 100 cm. Maciso (2017) — Oil on canvas, 100 × 100 cm. Y, Bueno (2017) — Oil on canvas, 80 × 80 cm. Redon\u0026rsquo;s Buddha (2017) — Acrylic on canvas, 20 x 30 cm. Monochrome (2017) — Oil on canvas, 50 × 100 cm. Threads series. Threads Series (2013-2022) # Works on canvas and paper tracing line, labor, and the wovenness of bodies.\nGimnasta (2018) — Acrylic on paper, 150 × 200 cm. Untitled (2013) — Acrylic on canvas, 50 × 40 cm. Integration (2017) — Acrylic on canvas, 40 × 50 cm. La Sagrada Familia (2017) — Acrylic on canvas, 80 × 100 cm. Delantales (2022) — Acrylic on canvas, 30 × 30 cm. ","externalUrl":null,"permalink":"/art/painting/","section":"Art","summary":" San Giorgio Blue [Crack Haha] (2022) — Acrylic on canvas, 40 × 30 cm., with Elisa Santos. (Non) Ho Pelli Sulla (2019) — Acrylic on canvas, 70 × 50 cm. Dalí (2017) — Acrylic on canvas, 100 × 120 cm. Ciggie Break (2021) — Acrylic on canvas, 30 × 30 cm. Chico Blue (2026) — Acrylic on canvas, 40 × 50 cm. Bearing Body Series (2017), oil on canvas. Bearing Body Series (2017) # Four-part series that isolates fragments of the bearing (the feminized as potentially pregnant) figure caught mid-gesture, attending to the body as a site of weight, posture, and the quiet labor of holding a/one self together.\n","title":"PAINTING","type":"art"},{"content":" \u0026gt; DETECTING SENSIBILITY... [CRITICO-COMPUTATIONAL] \u0026gt; CORE INQUIRY AXES... [IDENTITY] [TOXICITY] [VULNERABILITY] \u0026gt; FOLLOW THE GLITCH_ My research operates at the intersection of critical data studies, software auditing, and queer and media theory. I investigate how abstract, context-dependent cultural concepts—most centrally identity, toxicity, and vulnerability—are operationalized, classified, and flattened within datasets, models, generative AI architectures, and platforms.\nDuring my doctoral research, I investigated how pre-generative computer vision models were trained to \u0026ldquo;see\u0026rdquo; and measure abstract cultural concepts like freedom, comfort, danger, and power. This work surfaced a fundamental computational tension: the structural requirement of machine learning to reduce multifaceted, subjective human phenomena into static, discrete mathematical variables. Today, I extend this critique into the inner workings of large language and multimodal models, analyzing how this flattening operates across their latent representations and downstream platforms.\nRather than approaching classification errors, representational harms, or hallucinations as isolated bugs to be patched, my methodology centers on what I have coined as “following the glitch.” Drawing on glitch feminism, queer failure, and critical AI studies, I treat the moments where computational taxonomies and binary schemas break down as diagnostic windows. These glitches reveal how systems operate underneath their interfaces, while pointing to tactical sites for queer refusal, epistemic opacity, and counter-hegemonic design.\nConceptual Web # // CRITICO-COMPUTATIONAL SENSIBILITY Technofeminism • Queer Theory • Anticolonial Epistemologies • STS // OPERATIONALIZATION OF ABSTRACT CONCEPTS Taxonomic Flattening • Latency • Extraction • Algorithmic Interpellation Transforming fluid cultural phenomena into discrete, computable variables \u003e IDENTITY • Marked vs. Unmarked Norms • Pre-training Adult Taxonomies • Latent Visual Grammars • Hallucinated Subjects (\"You\"/\"I\") QueerGen • Latent Image • NSFW \u003e TOXICITY • Target–Intent–Tactic Triad • Multimodal Failure Modes • Memetic Trauma Extractivism • Content Moderation Taxonomies Toxic Memes • Memetic Journey \u003e VULNERABILITY • The \"Protection Paradox\" • Pipeline-Induced Precarization • Coercive Coherence in GenAI • Secondary Exposure \u0026 Fixing FAccT '26 • Hallucinated Subjects // METHODOLOGICAL STANCE: \"FOLLOWING THE GLITCH\" Empirical Audits • Deconstructing Classifiers • Strategic Opacity \u0026 Queer Ungovernability Treating systemic breakdown as a diagnostic tool rather than an engineering error For the complete publication list, see the Publications page and/or Google Scholar.\n","externalUrl":null,"permalink":"/research/","section":"Broken Binaries","summary":" \u003e DETECTING SENSIBILITY... [CRITICO-COMPUTATIONAL] \u003e CORE INQUIRY AXES... [IDENTITY] [TOXICITY] [VULNERABILITY] \u003e FOLLOW THE GLITCH_ My research operates at the intersection of critical data studies, software auditing, and queer and media theory. I investigate how abstract, context-dependent cultural concepts—most centrally identity, toxicity, and vulnerability—are operationalized, classified, and flattened within datasets, models, generative AI architectures, and platforms.\n","title":"RESEARCH","type":"page"},{"content":" RESEARCH # My work operates at the intersection of critical data studies, software development and auditing, queer and media studies, and anticolonial epistemologies. I study how abstract, context-dependent, and inherently subjective cultural concepts—most centrally identity, toxicity, and vulnerability—are operationalized, classified, and flattened within datasets, machine learning models, and generative AI architectures.\nDuring my doctoral research, I resaerched how pre-generative computer vision systems are taught to \u0026ldquo;see\u0026rdquo; and measure abstract cultural values like freedom, comfort, danger, and power. This revealed a fundamental computational tension: the structural necessity of machine learning to reduce multifaceted, culturally situated human phenomena into static, legible, and discrete mathematical variables. Today, I extend this critique to large language and vision-language architectures, analyzing how this flattening operates across their latent representations and downstream platform ecosystems.\nRather than approaching classification errors, representational harms, or hallucinations as isolated bugs to be corrected with larger datasets or parity benchmarks, my methodology centers on “following the glitch.” Drawing on glitch feminism, queer approaches to failure, and critical AI studies, I treat the moments where computational taxonomies, binary schemas, and colonial hierarchies break down as diagnostic and productive windows. These glitches reveal how systems operate underneath their interfaces, while exposing tactical sites for queer refusal, epistemic opacity, and counter-hegemonic practice.\nConceptual Web # // CRITICO-COMPUTATIONAL FOUNDATIONS Technofeminism (Preciado) • Queer Theory (Butler) • Opacity (Glissant) • Decolonial STS // OPERATIONALIZATION OF ABSTRACT CONCEPTS Classification • Taxonomic Compression • Latent Space Flattening Tracing how lived, context-dependent human realities become computable variables ⚯ BRIDGE: The Hallucinated Subjects \u003e IDENTITY • Binary \u0026 Taxonomic Modeling • Marked vs. Unmarked Norms • Latent Visual Grammars • Datafied \"You\" / Machinic \"I\" QueerGen Echoes of NSFW [WIP] Mapping Latent Image [WIP] \u003e TOXICITY • Target–Intent–Tactic Triad • Multimodal Failure Modes • Memetic Trauma Extractivism • Implicit Identity Target-Bias 'Toxic' Memes Survey Tracing a Memetic Journey \u003e VULNERABILITY • The \"Protection Paradox\" • Relational Precarization • Coercive Coherence in GenAI • Exposure \u0026 Narrative Fixing Vulnerabilizing Practices The Hallucinated Subjects // METHOD: \"FOLLOWING THE GLITCH\" Empirical Auditing • Deconstruction of Annotation Schemas • Ontological Critique Locating system breakdowns to mobilize Queer Opacity, Fugitivity \u0026 Ungovernability Core Research Inquiries # 1. Identity: Flattening, Latency, and the Hallucinated Subject # Computational systems require identity to be persistent, discrete, and legible. This research thread interrogates how gender, sexuality, race, and geography are modeled, compressed, and negotiated across LLMs, generative vision models, web-scale training data, and conversational interfaces.\nFLUID ONTOLOGY Situated \u0026 non-binary lived experience EXTRACTION Pornographic \u0026 web tags ⚑ Echoes of NSFW [WIP] LATENT MAPPING Normative defaults \u0026 pose ⚑ QueerGen • Latent Image [WIP] COERCIVE COHERENCE Hallucinated \"You\" \u0026 \"I\" ⚑ Hallucinated Subjects // GLITCH: Queer Opacity, Trans Fugitivity \u0026 Resistance Sentence Completion \u0026amp; Normative Baselines: Auditing how language models encode societal norms. Rather than relying on simplistic binary evaluations (queer vs. non-queer), this work audits model generations across four metrics (sentiment, regard, toxicity, and prediction diversity), demonstrating that models penalize queer-marked subjects while treating unmarked categories as the normative cis-heteronormative default.\n↳ Explore in QueerGen. Pre-training Corpora \u0026amp; Colonial Taxonomies: Auditing the persistence of commercial adult content within Common Crawl pre-training corpora, revealing that foundational models inherit explicit demographic taxonomies (Asian, Latina, Brazilian, Ebony) derived from pornographic platforms, embedding colonial hierarchies into baseline AI representations.\n↳ Ongoing work in collaboration with Goethe University Frankfurt: Echoes of NSFW [WIP]. Visual Grammars in Synthetic Portraiture: Auditing over 1,200 synthetic portraits generated from neutral demographic prompts, identifying 157 unprompted aesthetic features (gaze, pose, lighting, atmosphere) that position marginalized bodies within colonial and affective regimes of docility and vulnerability.\n↳ Ongoing collaboration with Affect Lab: Mapping the Latent Image [WIP]. Ontological Subject Fabrication: Theorizing AI hallucination not as an epistemic error, but as an infrastructure of subjectivation. Traces the recursive stabilization between the Hallucinated Human Subject (the datafied double built to be governable) and the Hallucinated Machinic Subject (the synthetic persona performing unrefusing care while obscuring the precarious, racialized data labor sustaining its existence).\n↳ Read the theoretical framework in The Hallucinated Subjects. 2. Toxicity: Multimodal Complexity, Taxonomies, and Trauma Extractivism # Automated moderation tools routinely reduce online harm to a binary classification task. This thread dissects the computational operationalization of \u0026ldquo;toxicity,\u0026rdquo; uncovering how benchmarks conflate distinct rhetorical phenomena and how platforms enable the extraction and circulation of historical violence.\n\"TOXICITY\" Conflated Umbrella Term TARGET Individual • Group • Society INTENT Harm • Disinform • Exploit TACTIC Slurs • Fallacies • Roles // MEMETIC TRAUMA EXTRACTIVISM 1. Historical State Terror (Global South) ▼ Platform Gamification \u0026 Subcultural Recoding 2. Alt-Right Identity Token (Global North) ⚑ Tracing a Memetic Journey Harmonizing Multimodal Toxicity: A systematic survey of 158 computational studies and 34 datasets evaluating toxic memes. This work untangles widespread terminological confusion by formalizing a meta-model across three independent dimensions: Target: Who is attacked (Individual, Organization, Community, Society). Intent: Why the content was produced (Harm/Abuse, Disinformation, Exploitation). Conveyance Tactics: How harm is delivered (Attack types, persuasion fallacies, entity roles).\nCritique: The survey demonstrates that automated classifiers inadvertently freeze demographic schemas, relying on hard-coded identity taxonomies as target proxies to infer whether an attack has occurred.\n↳ Access the meta-taxonomy in ‘Toxic’ Memes: A Computational Survey. Memetic Trauma Extractivism: Tracing the digital lifecycle of South American state terror—specifically the vuelos de la muerte (death flights) carried out by Southern Cone military dictatorships—into the alt-right \u0026ldquo;Free Helicopter Rides\u0026rdquo; meme on 4chan\u0026rsquo;s /pol/. Combining ten years of computational archives with digital ethnography, this work articulates Memetic Trauma Extractivism: how platform infrastructures detach historical trauma from the Global South, sanitize its state-terror origins, and convert it into gamified ideological tokens used to forge reactionary identities in the Global North.\n↳ Read the analysis in Tracing a Memetic Journey. 3. Vulnerability: From Inherent Traits to Vulnerabilizing Practices # Rather than viewing vulnerability as an intrinsic demographic deficit belonging to specific \u0026ldquo;vulnerable groups,\u0026rdquo; this research examines how vulnerability is actively manufactured, amplified, and precarized through data practices and algorithmic interventions.\n1. DATASET DESIGN Scale \u0026 Consent Gaps 2. OPERATIONALIZATION Affect \u0026 Nudity Labels 3. INFERENCE API Leakage \u0026 Thresholds 4. DISSEMINATION Narrative Fixing \u0026 Stigma // THE PROTECTION PARADOX (FAccT '26) Well-intentioned AI interventions risk manufacturing exposure, narrative fixing \u0026 extraction The Protection Paradox: Grounded in an AI for Social Good (AI4SG) case study involving computer vision pipelines designed to detect children in monetized family vlogs for regulatory advocacy, this work demonstrates how technical efforts to protect subjects often subject them to intensified extraction, exposure, and control. Reflexive Pipeline Junctures: We formulate an operational ethics protocol analyzing how granular technical choices across four pipeline stages produce precarity: Dataset Design: Navigating sampling scale, attention-mirroring biases, and the false equivalence between platform accessibility and ethical consent. Operationalization: Critiquing how fluid human affect and caregiving are reduced to rigid labels (distress, nudity), imposing moralized classifications and biometric tracking on minors. Inference \u0026amp; Evaluation: Analyzing data sovereignty risks when querying third-party APIs, and showing how confidence thresholding pathologizes ordinary domestic life. Dissemination: Preventing narrative fixing (locking subjects into permanent roles of victimhood) and mitigating the risk of research artifacts turning into discovery tools for online harassment.\n↳ Review the protocol in From Vulnerable Data Subjects to Vulnerabilizing Data Practices. Methodological Stance: \u0026ldquo;Following the Glitch\u0026rdquo; # Across all three threads, my work connects technical audits with critical theory to interrogate points of system failure:\nBeyond Inclusion-as-Capture: Expanding dataset diversity or adding granular classification labels frequently extends the reach of computational surveillance. Inclusion within reductive taxonomic architectures functions as algorithmic capture. My work pinpoints the boundaries where systems cannot and should not classify. Reflexivity as World-Making: Treating technical choices—from tokenization and bounding boxes to loss functions and data dissemination formats—as active, ethically constitutive decisions that script social reality. Affirming Opacity \u0026amp; Fugitivity: Drawing on Édouard Glissant’s right to opacity and Marquis Bey’s trans fugitivity, glitches are treated not as bugs to patch, but as structural limits that protect human multiplicity from totalized datafication. \u0026gt; THREE CORE AXES: \u0026gt; [IDENTITY] [TOXICITY] [VULNERABILITY] \u0026gt; _ IDENTITY Breaking or collapsing the binary and taxonomic (colonial) hierarchies of identity categories. This includes gender and sexuality binarisms or taxonomic ideas (male/female, queer/non-queer) and moving toward more complex framings like marked vs. unmarked.\nThis work is seen in the latest QueerGen article, my role as co-founder of the Queer and Feminist Informatics Network (QFIN), and two upcoming works:\n\"Echoes of NSFW\" — on the identity categories (demographic, national, regional, and racialized terms like asian, latina, brazilian, ebony, etc.) in NSFW pages that are part of the training data of foundational models through Common Crawl. This work is led by PhD student Ella Streefkerk at Goethe University Frankfurt. Synthetic portraiture — how identity/demographic labels like age, gender, and racial markers affect outputs in text-to-image models. This interest is centered on synthetic portraiture and developed in collaboration with the Affect Lab, currently finishing a paper on \"Mapping the Latent Image: Analyzing Representational Power and Harm in Synthetic Portraits.\" Recent publication: \"The Hallucinated Subjects: Gender, Coherence, and the Ontological Politics of Generative AI\" (with Siân J M Brooke, 2026) — develops a queer and technofeminist account of how GenAI hallucination operates as ontological subject production, fabricating gendered coherence through the hallucinated human subject and the hallucinated machinic subject.\n\u0026gt; STATUS: ACTIVE \u0026gt; PUBLICATIONS: QUEERGEN, QFIN, THE HALLUCINATED SUBJECTS \u0026gt; UPCOMING: ECHOES OF NSFW, MAPPING THE LATENT IMAGE \u0026gt; _ TOXICITY Questioning who and how decides who is toxic. Operationalizing abstract social concepts—toxicity, moderation, harm—in datafied environments.\nThis includes developing AI models for detecting and analyzing toxicity in online platforms, with a focus on identifying patterns of toxic symbology in memes. It also involves advocating for digital literacy skills and interrogating how morality and normativity are operationalized in content moderation systems.\n\u0026gt; STATUS: ACTIVE \u0026gt; FOCUS: MEMES / CONTENT MODERATION / DIGITAL LITERACY \u0026gt; _ VULNERABILITY Questioning who gets labeled vulnerable, and how that labeling becomes embedded in governance systems and surveillance practices.\nThis includes work on vulnerable data subjects—such as children in family vlogs—and the protection paradox in AI-based analyses of platformized lives. It also examines how vulnerability is co-constructed through AI systems and data practices, balancing public interest with individual rights.\n\u0026gt; STATUS: ACTIVE \u0026gt; FOCUS: CHILDFLUENCERS / DATA GOVERNANCE / PROTECTION PARADOX \u0026gt; _ \u0026gt; FULL PUBLICATION LIST: [PUBLICATIONS PAGE] \u0026gt; GOOGLE SCHOLAR: [LINK] \u0026gt; _ For the complete publication list, see the Publications page.\nGOOGLE SCHOLAR \u0026gt;\n","externalUrl":null,"permalink":"/research_old/","section":"Broken Binaries","summary":"RESEARCH # My work operates at the intersection of critical data studies, software development and auditing, queer and media studies, and anticolonial epistemologies. I study how abstract, context-dependent, and inherently subjective cultural concepts—most centrally identity, toxicity, and vulnerability—are operationalized, classified, and flattened within datasets, machine learning models, and generative AI architectures.\n","title":"RESEARCH","type":"page"},{"content":"","externalUrl":null,"permalink":"/tags/","section":"Tags","summary":"","title":"Tags","type":"tags"}]