Know Your Unknown Environmental Impact: The Influence of Eco-Feedback on AI Use during an Image Creation Task
Injin Park · 2026
With the advent of Artificial Intelligence (AI), the discussion of its energy demand has gained attention, given its high frequency of use. However, discussions have been heavily focused on the training phase, and the inference and usage side have often been overlooked. Thus, we explored AI user behavior, especially in the image-creation task, which is known to be among the most energy-intensive tasks that AI can perform. As a solution, we devised novel interfaces, where one focused on informing the short-term energy impact of AI usage (Short-Term Impact), and another focused on the long-term energy impact (Long-Term Impact). In particular, we investigated whether interfaces encourage users to use them in an environmentally conscious manner. We hypothesized that both interfaces would: (1) significantly reduce the number of iterations, (2) significantly increase average hypothetical donations, (3) significantly increase policy support, and (4) significantly improve awareness of AI’s energy consumption compared to the control condition (No Impact). A pre-registered experiment via online survey (N = 300) showed that the Short-Term Impact condition significantly reduced the number of iterations compared to the No Impact condition, but did not find any significant results in all other aspects. We also did not find any significant results from the Long-Term Impact condition. This research provides a groundwork for how AI users behave when they are given the user interface (UI)-based interventions. This attempt will encourage future studies for devising new interface designs to create a greater impact on the positive behavioral change for our sustainable future.