(A)I CAN’T SEE HER

Abstract

Histobots (Harder 2024b), AI-driven chatbots that simulate historical figures, are marketed as tools for education and engagement. They promise immersion but operate within a system of algorithmic mediation that flattens complexity and reinforces dominant narratives. Unlike traditional historical interpretation, which involves deliberate source selection and critical framing, histobots generate responses based on probabilistic patterns, presenting history as seamless, neutral, and objective. This illusion of neutrality conceals deep biases embedded in training data, filtering mechanisms, and corporate imperatives. This paper examines histobots as algorithmic reenactments. It explores how AI reshapes historiography through feminist and queer theoretical lenses. I reflect on my experience developing a histobot of Hedy Lamarr and analyse AI-generated representations of figures such as Anne Frank, Harriet Tubman, and Marsha P. Johnson. These chatbots erase political agency, neutralise rhetorical power, and homogenise voices. They produce a form of historical negationism that tokenises rather than represents marginalised figures. I draw on situated knowledges (Haraway 1988) and critical AI scholarship (Crawford 2021; Felkner et al. 2024) and argue that histobots reproduce epistemic injustices by encoding archival silences and structural biases. They inherit the exclusions of the historical record while reinforcing contemporary inequalities. This paper interrogates whether histobots can be reclaimed for feminist storytelling or whether, as Audre Lorde (1979) cautions, “the master’s tools will never dismantle the master’s house.” Can AI-driven history ever be ethical, or must we build new tools entirely?