Generative AI’s Family Portraits of Whiteness: A Postdigital Semiotic Case Study

While traditional family portraits have always been value-laden constructs, generative artificial intelligence’s text-to-image tools are adding new postdigital complexities to already sophisticated photo and video editing processes. To explore what is being produced or generated by text-to-image generative artificial intelligence, this postdigital semiotic case study assembles a comparative corpus of family portraits. Analysis of portraits from three different platforms indicates that artificial intelligence is not entirely devoid of agency but rather a type of language-user entangled within complex processes of technosemiosis resulting in a series of hallucinations. Theorising postulates at least three distinct types of visual hallucinations, including hallucinations as perceived errors, hallucinations as a distorted view of reality, and hallucinations as regimes of Whiteness. In view of the ethnocentric ‘myths’ emerging from the corpus of hallucinations, the study articulates ethical imperatives for developing human-in-the-loop guardrails. In the process, the study debunks not merely the seeming ‘magic’ of artificial intelligence’s algorithmic black box but also the quick fixes of predetermined theoretical approaches. As a result, the study raises broader issues concerning the openness of postdigital research which is required to address the multifaceted contagion of racial capitalism. While navigating these opaque constellations, the study reinforces the axiological scope of the postdigital semiotic framework to open theorising beyond scholarly performativity and epistemic complacency.