Title Pending

When algorithms speak in the voices of the dead, who decides what they say? This essay examines that question through a situated and at times uncomfortable reflection on the design and failure of a prototype of what the author terms a histobot: a conversational agent that uses generative Artificial Intelligence to imitate a person from the past. The case study is a histobot of Hedy Lamarr, the Austrian-American actress and inventor, which the author constructed in 2024 using OpenAI's GPT-4o mini. What began as a technical exercise became an inquiry into the assumptions and infrastructures that shape AI-mediated historical simulation. No matter how carefully the system was instructed, it could not hold the situatedness of Lamarr's life. Drawing on Donna Haraway's situated knowledges and Shannon Vallor's concept of AI as a mirror, and informed by feminist, decolonial, and intersectional critiques of algorithmic systems, the essay argues that to simulate a person is to interpret them, and to automate that simulation is to encode that interpretation into an opaque system that mistakes dominant historiography for historical truth. It traces how WEIRD training data, platform alignment layers, and archival silences compound one another, disproportionately erasing marginalised voices across histobot platforms, and considers what more accountable practice might look like.