Generative AI (genAI) can produce cultural artefacts at scale, yet its design also embeds prevailing cultural values. Once these values are identified, they become open to intentional reshaping. This paper adopts an environmental humanities perspective to critique the maximalist values of current genAI and proposes a sustainability‑centered design framework called Slow AI.
Slow AI articulates five design principles: restraint, sufficiency, selectivity over retention, material visibility, and friction as affordance. Each principle offers a concrete implementation suggestion and an interpretive layer that prompts users and developers toward reflective engagement. For instance, restraint advises default limits on generation volume, lifted only upon explicit request; sufficiency encourages models to meet task needs with minimal resource use; selectivity over retention provides visual tools for users to decide which outputs to keep; material visibility makes energy consumption and data provenance transparent; friction as affordance inserts deliberate confirmation steps to curb mindless generation.
Operating on both technical and interpretive levels, these principles restore decision points removed by frictionless defaults and embed reflection into the user experience, thereby extending human agency. By foregrounding environmental sustainability, Slow AI seeks a more balanced relationship between cultural production and resource consumption.
Blogger's Review: The Slow AI approach offers a pragmatic ethical scaffold for generative models, and its principles merit incremental testing in real‑world systems.