How can feminist concepts like 'performativity' transform the way we think about, create, and legislate AI?
With Dr Federica Frabetti I have been applying Karen Barad and Judith Butler's concepts of performativity to critical issues in AI, from racist policing software to biometric technologies used in Germany to validate migration claims.
See the Projects page for my TikTok videos on protest recognitions software, commissioned by Carole Cadwalladr's The Citizens
Frabetti's phenomenal cultural and philosophical study of software, which explores how code is constitutive of culture and even of human thought.
Federica and I have written three papers on how AI can be described as performative and why we urgently need to understand its behaviour as such
1) "AI that Matters: A Feminist Approach to the Study of Intelligent Machines"
In this chapter in Feminist AI, forthcoming with Oxford University Press, we introduce performativity, a well-established concept in gender studies and feminist science studies, as a useful tool for explaining in detail how AI produces the effects of gendered embodiment that it claims to describe or ‘identify’.
Feminist thinkers have played an important role in exposing and communicating AI’s integration within and exacerbation of gendered power systems. Expanding on this, we argue that performativity, a well-established concept in gender studies and feminist science studies, is a useful tool for explaining in detail how AI produces the effects of gendered embodiment that it claims to describe or ‘identify’. Theories of performativity have revealed, for example, that a person’s gender is not an innate or binary human attribute but is instead enacted through repeated behaviours and activities, as Judith Butler proposed in the 1990s. The concept of performativity has also been used to insist that the material world does not precede our scientific and cultural observations of it but is actively produced by them, a phenomenon which Karen Barad calls ‘agential realism’. Without downplaying their differences, we stage a dialogue between Butler’s and Barad’s concepts of performativity in relation to AI to explain how AI creates the effects that it names. We begin with an explanation of Butler’s and Barad’s theories and what kind of work they do in the study of software and AI, before locating AI’s performative origins in Alan Turing’s imitation game, in which gender is the modality through which the thinking computer can be perceived as such. Next, we show how both Butler’s and Barad’s concepts of performativity work together in applying feminist knowledge to AI at the level of its technical and conceptual functioning. Finally, we engage in a close reading of neural networks in the context of Facial Detection and Recognition Technologies (FDTR) and Automatic Gender Recognition (AGR). We show how these technologies make claims about the world that generate gendered and racialised interpretations of bodies in accordance with hegemonic value systems by iteratively ‘citing’ social norms in the act of ‘observing’. This work translates concepts from feminist and gender studies into meaningful ways forward for AI practitioners, users (‘participants’1) and stakeholders when responding to the issue of AI’s implication in existing power structures.
2) "Copies without an Original: the Performativity of Biometric Bordering Technologies"
In this article forthcoming with Communication and Critical/Cultural Studies, we analyze two examples of biometrics in civil registration and migration contexts (the German Federal Office for Migration and Refugees’ voice biometry system and UK HMPO passport photo checker tool).
We use the concept of performativity to show how, rather than ‘recognizing’ a person, biometrics creates a field of intelligibility within which the shifting positionalities of bodies are ‘stabilized’ and deemed recognisable. We show how the obfuscation of this process has had violent racializing and gendering effects on the bodies of AI participants. We present our performative approach as a strategic intervention at the intersections of AI ethics and biometrics in these contexts.
3) "The Performativity of AI-Powered Event Detection: How AI Creates a Racialized Protest and Why Looking for ‘Bias’ Is Not a Solution"(forthcoming in Society, Technology & Human Values)
In this article forthcoming with Theory, Culture & Human Values, we use the concept of performativity to confront the racializing capabilities of AI-powered real-time Event Detection and Alert Creation (EDAC) software when used for protest detection.
It is well known that many AI-powered systems exacerbate social inequalities by racializing certain groups and individuals. We propose the feminist concept of performativity, as defined by Judith Butler and Karen Barad, as a more comprehensive way to expose and contest the harms wrought by EDAC than other ‘de-biasing’ mechanisms. We explain how our use of performativity differs from and complements other STS work because of its rigorous approach to how iterative, citational, and material practices produce the effect of race. We focus on Geofeedia and Dataminr, two EDAC companies that claim to be able to ‘predict’ and ‘recognize’ the emergence of dangerous protests, to show how their tools performatively produce the phenomena which they are supposed to observe. Specifically, we argue that this occurs because these companies and their stakeholders dictate the thresholds of (un)intelligibility, (ab)normality and (un)certainty by which these tools operate, and that this process is oriented towards the production of commercially actionable information.