Bogdan-Andrei Lungu

This paper examines the transformations of personhood in the digital age brought by the recursive operations of machine learning (ML) artificial intelligence systems (AI). Focusing on the opaque ways recursive machine learning systems construct specific “digital human twins” (DHTs) as representations of real persons, it analyzes how contemporary algorithmic infrastructures entangle human selfhood and how this, in turn, impacts autonomy, agency, and self-determination. Through a case study conducted on mental health chatbots, this paper showcases how ML feedback-based loops operationalize and often constrain the complexity of human identity, rendering individuals as datafied entities caught within opaque recursive technological systems. This analysis proposes a new concept called the “recursivisation of personhood” to aid analytical investigations into the shifts brought by ML/AI ecology on contemporary notions of digital identity. This paper showcases that personhood online has become entangled in data extractive recursive processes with the advent of ML/AI systems. In response to the possible dangers identified from this socio-technical development, this paper proposes an ethics of “de-looping”—a new normative framework aimed at interrupting looped ML operations and practices and rendering transparent the recursive processes that form digital human twins. Arguing for the need of a new “infra-ethical” infrastructure of de-looping as a means of making the production of digital human twins transparent, this work contributes to ongoing debates about identity online, digital rights, algorithmic governance, and the future of human agency in an age increasingly shaped by ML/AI systems.