The Creative Ghost in the Algorithm: AI-Generated Photo-Based Images
Forthcoming.
Abstract
This paper explores the ontological and pragmatic status of AI-generated photo-based images, situating them within ongoing philosophical debates on photographic indexicality, authorship, and visual deception. I argue that AI-generated photo-based images – those produced by diffusion models trained on photographic data – should be understood as quasi-indexical images. While lacking the direct causal connection to the world that defines traditional straight photographs, they retain enough formal and stylistic continuity with photographic conventions to participate in many of the same communicative practices. This quasi-indexicality places them ontologically between fully indexical photographs and non-indexical media such as painting or drawing. Notably, this in-between status is not unique to AI images; it also characterizes other photographic hybrids such as multiple exposures and photo collages, which similarly disrupt one-to-one reference while maintaining partial indexical grounding. The production of these images also exhibits the kind of glitch process that has gained considerable popularity in photography, particularly in lomography.
This ontological framing informs a broader inquiry into how AI-generated images function in communicative contexts. Building on works on photographic illocutionary acts and deception, I examine whether and how these images can perform pictorial illocutionary acts – asserting, warning, promising – despite their lack of indexical grounding. While traditional photographic illocutions rely on an implicit trust in the image’s connection to reality, AI-generated images may mimic this force without its epistemic foundation. This raises critical questions: Can AI-generated images “pretend” to assert? Do they deceive by exploiting viewers’ residual expectations of photographic veracity? Or do they give rise to a distinct kind of illocution requiring new conceptual categories within visual pragmatics?
These questions are pursued along three interrelated lines of inquiry: (1) the ontology and authorship of AI-generated photo-based images, with implications for copyright, creative agency, and the classification of hybrid media; (2) the creativity involved in prompt design and model interaction, exploring how human agency is exercised in collaborative image generation; and (3) the pragmatics of deception and interpretation, including empirical research on how audiences perceive the communicative intent and truth-status of AI-generated images.
Ultimately, this project argues that AI-generated photo-based images constitute a new subclass within photo-based representation – quasi-indexical images that challenge our assumptions about visual realism, authorship, and communicative intent. By focusing on the specific features of this emerging visual paradigm, the paper contributes to medium-specific approaches to generative AI, bridging philosophy of photography, aesthetics, media theory, and epistemology.
Bibliography:
Epstein, Z. et al. (2023a). Art and the science of generative AI: A deeper dive. arXiv:2306.04141 [cs.AI], 1–23. doi: https://doi.org/10.48550/arXiv.2306.04141
Epstein, Z. et al. (2023b). Art and the science of generative AI: Understanding shifts in creative work will help guide AI’s impact on the media ecosystem. Science, 380, 1110–1111. doi: https://doi.org/10.1126/science.adh4451
Kjørup, S. (1978). Pictorial Speech Acts. Erkenntnis, 12, 55-71.
Kjørup, S. (1974). George Inness and the Battle at Hastings, or Doing Things With Pictures. The Monist, 58:2, 216-235.
Novitz, D. (1977). Pictures and their Use in Communication: A Philosophical Essay. The Hague: Martinus Nijhoff.