Slow AI Proposes Sustainable Design Principles for Generative Systems

Current generative AI systems, while producing cultural artifacts at scale, often embed maximalist cultural values in their design. This position paper argues for a deliberate reshaping of these values, proposing "Environmental Slow AI" as a framework centered on environmental sustainability. The paper articulates five design principles: restraint, sufficiency, selectivity over retention, material visibility, and friction as affordance. Each principle is illustrated by contrasting it with the current design of widely deployed generative AI systems. For example, "restraint" suggests limiting unnecessary generation, while "material visibility" encourages transparency about resource consumption. These principles operate on two levels: practical design implementation and an interpretive layer that prompts users and developers to reflect on their engagement with the system. By reintroducing decisions that frictionless defaults have removed, Slow AI aims to extend human agency and foster more reflective, environmentally conscious AI development and use.