When
Where
Title: Modern digital technology and the limits of human discrimination
Abstract: We are increasingly confronted with technologies and technological outputs that cannot be discriminated from their non-technological counterparts. For instance, deepfakes cannot be told apart from recordings, and large language models’ outputs cannot be told apart from human discourse. How should we think about the epistemic and moral impact of these technologies? In what follows, I argue that these technologies and others result in discriminability sabotage. Discriminability sabotage is an epistemic phenomenon that occurs when epistemic environments are modified in ways that result in agents losing the ability to discriminate in some domain they can ordinarily discriminate in. I discuss this phenomenon in four parts. First, I describe discriminability sabotage as part of a broader class of epistemic phenomena, environmental epistemic sabotage. Second, I explain how various digital technologies, including deepfakes, large language models, virtual and augmented reality, robotics, social media, and machine learning models, are discriminability sabotaging. Third, I argue that along with the epistemic threat, discriminability sabotage can also pose ethical and even existential threats. Finally, I propose and assess three ways of avoiding the threats posed by discriminability sabotage.
As usual, we'll meet in the Maloney Seminar Room, Social Science Building 224, 3-5p. Those unable to attend in person can spectate virtually via this Zoom link