Researchers Say Simple Design Changes Can Encourage Greener AI Use

Researchers Say Simple Design Changes Can Encourage Greener AI Use

A new study from Oregon State University suggests that small design tweaks in artificial intelligence systems could help users make more environmentally responsible decisions when using AI tools.

Published in the journal Science Communication, the research found that prompting users to slow down and reflect on AI’s environmental impact can reduce unnecessary AI usage, particularly when generating images.

The study was led by Cheng “Chris” Chen from OSU’s School of Communication, alongside collaborators from the University of Illinois and the University of Virginia.

Researchers examined the effects of “design friction” — small interruptions or extra steps added to AI systems. One example required users to first search for existing image resources and provide detailed prompts before generating new AI images. According to the study, this “action-based friction” encouraged users to think more carefully about their AI consumption and act more responsibly.

The team also tested “cue-based friction,” which involved displaying messages about AI’s environmental costs. While these messages increased user trust, they had less impact on changing user behavior.

Chen explained that most AI systems are designed to maximize convenience and engagement, often encouraging users to keep interacting with the platform even after completing their original task.

The researchers highlighted the growing environmental concerns surrounding AI, noting that training a large language model can consume enough electricity to power around 120 homes for a year. They also pointed out that generating a single AI image may use about the same amount of energy as charging a smartphone.

With fossil fuels still supplying roughly 85% of global energy, the researchers say reducing AI-related electricity use is becoming increasingly important as AI adoption rapidly expands worldwide.

The study also offered practical tips for responsible AI use, including avoiding unnecessary AI-generated content, minimizing duplicate AI projects, and closing AI tools once tasks are complete instead of continuing prolonged interactions.

Researchers warn that high-performance computing and AI systems could account for up to 20% of global energy consumption by 2030 if usage continues to accelerate.