Bias as a signal: a cybernetic view on target states in AI

AIKG-SD 2025 Summer School co-located with the NFDI4DS Conference 2025 November 25-26, 2025, Berlin, Germany 1 Bias as a signal: a cybernetic view on target states in AI Anna Ehrenberg Abstract In a cybernetic framework of AI-society interactions, bias can be understood as a deviation between a system’s current state and the target state it is designed to reach. This poster explores the question: How can such target states be practically defined for tool-specific and demographically diverse user groups? Through a synthesis of literature and conceptual inquiry, we map the epistemological, technical, and ethical challenges involved in defining and negotiating target states within AI systems. Cybernetics provides a way to model AI development and societal impact as interlinked feedback loops. This perspective makes visible how system goals emerge not as objective facts but as socially constructed choices shaped by developer assumptions, institutional incentives, regulatory pressures, and historical social structures. As AI systems reach broader audiences and globalization progresses, the difficulty of defining shared target states grows, especially in the presence of conflicting values such as efficiency, equity, and autonomy. The poster outlines several unresolved tensions. Epistemological challenges arise from situated knowledge: dominant social groups influence what is perceived as a problem and whose experiences are reflected in system design. Technical challenges include translating complex values into measurable parameters, relying on imperfect proxies, and balancing transparency with competitive constraints. Ethical challenges surface around context-setting, prioritizing particular user groups, and determining whose needs and futures are represented in design choices. These challenges show how attempts to define “desirable” system states are intertwined with power, culture, and structural inequities. Rather than offering definitive solutions, this poster frames bias as a signal within a cybernetic system. It shows an indication of misalignment between chosen goals and the lived realities of affected communities. By foregrounding these unresolved tensions, the poster aims to support more nuanced discussions about how target states in AI are constructed, contested, and continuously reshaped through feedback between technology and society. References Anderson, T. and Shepherd, P. (2025) ‘Modeling and simulation of human societies: from challenges to new opportunities’, SIMULATION, 101(6), pp. 623–623. Available at: https://doi.org/10.1177/00375497251343097. Arora, P. and Chowdhury, R. (2021) ‘Cross-Cultural Feminist Technologies’, Global Perspectives, 2(1), p. 25207. Available at: https://doi.org/10.1525/gp.2021.25207. Ashby, W.R. (1956) ‘An introduction to cybernetics’. Available at: https://philpapers.org/archive/ASHAIT (Accessed: 23 October 2025). Azzopardi, L. (2021) ‘Cognitive Biases in Search: A Review and Reflection of Cognitive Biases in Information Retrieval’, in Proceedings of the 2021 Conference on Human AIKG-SD 2025 Summer School co-located with the NFDI4DS Conference 2025 November 25-26, 2025, Berlin, Germany 2 Information Interaction and Retrieval. CHIIR ’21: ACM SIGIR Conference on Human Information Interaction and Retrieval, Canberra ACT Australia: ACM, pp. 27–37. Available at: https://doi.org/10.1145/3406522.3446023. Beer, S. (2002) ‘What is cybernetics?’, Kybernetes, 31(2), pp. 209–219. Available at: https://doi.org/10.1108/03684920210417283. Cody, T., Adams, S. and Beling, P. (2020) ‘Motivating a Systems Theory of AI’, INSIGHT, 23(1), pp. 37–40. Available at: https://doi.org/10.1002/inst.12283. D’Ignazio, C. and Klein, L.F. (2020) Data Feminism. The MIT Press. Available at: https://doi.org/10.7551/mitpress/11805.001.0001. Drage, E., McInerney, K. and Browne, J. (2024) ‘Engineers on responsibility: feminist approaches to who’s responsible for ethical AI’, Ethics and Information Technology, 26(1), p. 4. Available at: https://doi.org/10.1007/s10676-023-09739-1. Feminist Internet and Young, J. (2020) ‘Feminist Design Tool Defensible Decision Making for Interaction Design and AI’. Available at: https://ugc.futurelearn.com/uploads/files/16/b0/16b088ad-6145-45eb-b5d83753a41b4b88/2-10_FeministDesignTool_2.0.pdf (Accessed: 12 November 2024). Ferguson, K.E. (2017) ‘Feminist Theory Today’, Annual Review of Political Science, 20(1), pp. 269–286. Available at: https://doi.org/10.1146/annurev-polisci-052715-111648. Garg, S. et al. (2019) ‘Counterfactual Fairness in Text Classification through Robustness’. arXiv. Available at: https://doi.org/10.48550/arXiv.1809.10610. Genero, D. (2024) ‘Feminisms in Artificial Intelligence: Automation Tools towards a Feminist Judiciary Reform in Argentina and Mexico | AymurAI’, Feminist AI [Preprint]. Available at: https://feministai.pubpub.org/pub/z83eyq54/release/1 (Accessed: 4 October 2024). Gengler, E.J. et al. (2023) ‘Power Imbalances in Society and AI: On the Need to Expand the Feminist Approach’. Haraway, D. (1988) ‘Situated Knowledges: The Science Question in Feminism and the Privilege of Partial Perspective’, Feminist Studies, 14(3), pp. 575–599. Available at: https://doi.org/10.2307/3178066. Hayles, N.K. (2023) ‘Technosymbiosis: Figuring (Out) Our Relations to AI’, in J. Browne et al. (eds) Feminist AI: Critical Perspectives on Algorithms, Data, and Intelligent Machines. Oxford University Press, p. 0. Available at: https://doi.org/10.1093/oso/9780192889898.003.0001. Heylighen, F. and Joslyn, C. (2003) ‘Cybernetics and Second-Order Cybernetics’, in Encyclopedia of Physical Science and Technology. Elsevier, pp. 155–169. Available at: https://doi.org/10.1016/B0-12-227410-5/00161-7. Klein, L. and D’Ignazio, C. (2024) ‘Data Feminism for AI’, in The 2024 ACM Conference on Fairness, Accountability, and Transparency. FAccT ’24: The 2024 ACM Conference on Fairness, Accountability, and Transparency, Rio de Janeiro Brazil: ACM, pp. 100– 112. Available at: https://doi.org/10.1145/3630106.3658543. Mühlhoff, R. (2025) ‘AI Cybernetics’, in The Ethics of AI. Bristol University Press, pp. 135– 145. Available at: https://bristoluniversitypressdigital.com/monochapoa/book/9781529249262/ch008.xml (Accessed: 23 October 2025). AIKG-SD 2025 Summer School co-located with the NFDI4DS Conference 2025 November 25-26, 2025, Berlin, Germany 3 Polcumpally, A.T. (2022) ‘Artificial intelligence and global power structure: understanding through Luhmann’s systems theory’, AI & SOCIETY, 37(4), pp. 1487–1503. Available at: https://doi.org/10.1007/s00146-021-01219-8. Purkayastha, B. (2012) ‘Intersectionality in a Transnational World’, Gender & Society, 26(1), pp. 55–66. Available at: https://doi.org/10.1177/0891243211426725. Radanliev, P. (2025) ‘AI Ethics: Integrating Transparency, Fairness, and Privacy in AI Development’, Applied Artificial Intelligence, 39(1), p. 2463722. Available at: https://doi.org/10.1080/08839514.2025.2463722. Srivastava, A. et al. (2023) ‘Beyond the Imitation Game: Quantifying and extrapolating the capabilities of language models’. arXiv. Available at: https://doi.org/10.48550/arXiv.2206.04615. Steen, M. et al. (2025) ‘We need better images of AI and better conversations about AI’, AI & SOCIETY, 40(5), pp. 3615–3626. Available at: https://doi.org/10.1007/s00146-02402101-z. Toupin, S. (2024) ‘Shaping feminist artificial intelligence’, New Media & Society, 26(1), pp. 580–595. Available at: https://doi.org/10.1177/14614448221150776. Wudel, Alexandra (2023) Navigating the Intersection: Disinformation, Deepfakes and Gender-Based Violence, FemAI. Available at: https://www.femai.com/post/navigating-the-intersection-deepfakes-disinformation-and-gender-basedviolence (Accessed: 10 October 2024).