“The Ethics of Simple Solutions” – Martin Essemann
It was difficult not to understand Zarzycka’s presentation in relation to Google’s decision to fire two of its top ethical AI researchers earlier this year. Zarzycka brought up this controversial decision herself in trying to elaborate on her own difficulties and feelings about becoming a Googler, but she is, understandably, not in a position to make sweeping public judgments on the matter. It would have been interesting to discuss Zarzycka’s views on the best application of ethics research in the development of Google’s products — as she currently seems to be involved in the form of post-hoc harm reduction and inclusion efforts that Gebru and Mitchell argued should be reintroduced as an integral part of the preliminary steps instead (Bender, Gebru, McMillan-Major, and Shmitchell 2021) – before they were fired.
The description of the precise nature of Zarzycka’s research and her findings were a bit too vague to elicit much well-founded criticism (I assume for confidentiality reasons). Instead, the discussion of ethics became a question of whether it is possible to conduct “ethical research” as part of a for-profit institution. But such a question takes for granted the idea that the academic freedom experienced in certain privileged institutions around the world is the norm, which of course is not the case — and particularly not in the for-profit university context of the United States, as Zarzycka reminded us.
One of the most interesting lessons shared by Zarzycka was the insight that outside decision-makers (whether they be product managers, funding committees, or public bodies) are much more likely to understand the importance of research into diversity and inclusion when it is presented as a way to avoid potential risks (harm reduction) rather than some more idealistic formulation of potential good. But such a formulation of discrimination and bias could fall into a simplistic mode of presenting “bias” as something that can be eliminated, provided enough resources are invested and sufficient critique carried out. This differs from the way many humanities scholars understand meaning and bias as closely connected. For philosophers like Hans-Georg Gadamer (2004) or Donald Davidson (1974) prejudices and systems of beliefs, respectively, are fundamental to our way of making sense of the world. Coming from such a perspective, the question of “dealing with bias” becomes rephrased as one of deciding what might constitute an acceptable, or at least tolerable, bias. Thinkers like Rosi Braidotti and Sylvia Wynter (2015) propose that we need to propose new “generative narratives” (Braidotti 2019) for how to deal with the current state of crisis — which is a quest that goes far beyond promoting accessibility for Google products. The question of ethics then becomes much more difficult but also pertinent: How do we communicate this complex approach to bias and ethics of the humanities to an audience that is only interested in clear-cut problems and simple solutions?
- Bender, Emily M., Timnit Gebru, Angelina McMillan-Major, and Shmargaret Shmitchell. 2021. “On the Dangers of Stochastic Parrots: Can Language Models Be Too Big? 🦜” In Proceedings of the 2021 ACM Conference on Fairness, Accountability, and Transparency, 610–623. FAccT ’21. Virtual Event, Canada: Association for Computing Machinery.
- Gadamer, Hans-Georg. 2004. “Elements of a Theory of Hermeneutic Experience.” In Truth and Method, Second, Revised Edition, 267–382. London, New York: Continuum.
- Davidson, Donald. 1974. “Belief and the Basis of Meaning.” Synthese 27 (3/4): 309-323.
- Sylvia, Wynter. 2015. “Unparalleled Catastrophe for Our Species? Or, to Give Humanness a Different Future: Conversations.” In Sylvia Wynter, On Being Human as Praxis, 9–89. Durham, London: Duke University Press.
- Braidotti, Rosi. 2019. “Posthuman Subjects.” In Posthuman Knowledge. Cambridge: Polity Press.