From Awareness to Action? CO₂ Feedback and Student LLM Use
The paper “From Awareness to Action? The Impact of CO₂ Emission Feedback on Student LLM Usage” was published in the ACM Digital Library and developed in collaboration with LMU Munich. Its central thesis is that the environmental impact of LLM use should not remain hidden in the background; when CO₂ feedback becomes visible at the moment of interaction, it can change how people use AI.
Generative AI often feels weightless. A prompt is written, a response appears, and the infrastructure behind it disappears from view. Yet every interaction has a material side: tokens processed, energy consumed, emissions produced. This project asks what happens when that material side becomes part of the learning experience.
In the study, students received real-time feedback on the estimated CO₂ emissions of their LLM interactions. Rather than presenting sustainability as a distant principle, the intervention brought it into the flow of use. Prompting, revising, repeating, and choosing how much to ask became moments in which environmental impact could be noticed and reconsidered.
The LMU collaboration is central to the project. It connects our work on responsible AI tools with empirical research on behavior, learning, and human-AI interaction. The question is not only whether students become aware of emissions, but whether awareness changes practice.
Initial findings suggest that real-time feedback can reduce emissions and total tokens per prompt, indicating that informational feedback may influence selected patterns of student LLM use. This matters because it shows that sustainable AI use does not have to rely only on infrastructure choices made elsewhere. It can also be shaped through design, feedback, and everyday interaction.
The project extends our work on the CO₂ Tracker and the responsible AI learning space. It turns sustainability into something users can see while they work with AI: not as a warning from outside the system, but as part of learning how to use the system more deliberately.