Description:

Synthetic Images, Synthetic Minds (SISMi) investigates how generative AI image systems are transforming images from discrete artefacts into cognitive interfaces that couple language, perception and decision-making. Anchored in visual culture, media and communication studies, STS and philosophy of mind, the project shifts attention from individual AI-generated pictures to the model pipelines, datasets and interfaces that structure what can be pictured, imagined and known. It examines how text–image–text loops, everyday prompting practices and platform defaults recalibrate people’s imaginative priors and habits of evaluation – a process the project terms “cognitive hacking” – and develops a framework for “image-thinking” that treats images as active instruments of reasoning rather than passive representations. Over four years, SISMi will generate an Image-Thinking Primer for teaching, a Model-as-Medium Field Guide, open interpretability notebooks, a media-literacy toolkit and a public exhibition that together equip scholars, students and wider publics to read synthetic image systems as cultural operators rather than opaque technical utilities. The project is led by Yanai Toister (PI) at Tampere University’s Visual Studies Lab, working with a postdoctoral researcher and doctoral researchers.

Seed article:

Yanai Toister and Joanna Zylinska. “Image Thinking after Artificial Intelligence”. Journal of Visual Culture (forthcoming 2025). (link) 

Technical description:

SISMi is structured into four interdependent work packages that combine conceptual analysis, ethnography and protocol studies with critical technical practice on diffusion models and related imaging pipelines. It uses protocol-led fieldwork with image practitioners, versioned model notebooks and contrastive “latent cartographies” to trace how upstream choices in datasets, embeddings, conditioning and safety layers shape the distribution of the visible. All work is implemented through open-science practices – a versioned web hub, reproducible code, bilingual teaching materials and public events – to ensure that its concepts, tools and demonstrations remain accessible, technology-durable and reusable beyond the project’s lifetime.

Project support: 

Seed support has been provided by the DigiSus platform at Tampere University (link)