Abstract
This paper examines how design research can unmake AI's classificatory ontologies through situated and embodied annotation practices. Drawing on feminist, queer, and critical data epistemologies, it reframes bias not as a technical flaw but as a generative condition that reveals how knowledge is produced and negotiated within human–AI assemblages. An autotheoretical experiment grounds the inquiry. The researcher creates and annotates a dataset of self-representations using subjective, affective, and relational labels, training a small-scale model to probe classification as an interpretive encounter shaped by embodiment, ambiguity, and positionality. Building on the shift from debiasing toward reflexive data practices in design and HCI, the paper proposes situated annotation as a design inquiry for unsettling inherited AI ontologies and repositioning machine vision as an accountable, partial, and embodied way of knowing.
Keywords
research through design, situated annotation, feminist epistemologies, human-AI assemblages
DOI
https://doi.org/10.21606/drs.2026.1239
Citation
Autuori, A. (2026) Unmaking AI’s classificatory ontologies: Situated annotation as design inquiry in human–AI assemblages, in Simeone, L., Gray, C. M., Verhoeven, A., de Götzen, A., Bakırlıoğlu, Y., Zohar, H., Stead, M., and Buwert, P. (eds.), DRS2026: Edinburgh, 8–12 June, Edinburgh, United Kingdom. https://doi.org/10.21606/drs.2026.1239
Creative Commons License
This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License
Included in
Unmaking AI’s classificatory ontologies: Situated annotation as design inquiry in human–AI assemblages
This paper examines how design research can unmake AI's classificatory ontologies through situated and embodied annotation practices. Drawing on feminist, queer, and critical data epistemologies, it reframes bias not as a technical flaw but as a generative condition that reveals how knowledge is produced and negotiated within human–AI assemblages. An autotheoretical experiment grounds the inquiry. The researcher creates and annotates a dataset of self-representations using subjective, affective, and relational labels, training a small-scale model to probe classification as an interpretive encounter shaped by embodiment, ambiguity, and positionality. Building on the shift from debiasing toward reflexive data practices in design and HCI, the paper proposes situated annotation as a design inquiry for unsettling inherited AI ontologies and repositioning machine vision as an accountable, partial, and embodied way of knowing.