Sound advice? The effect of an auditory prompt on recycling contamination

Recycling contamination creates economic and environmental costs, undermining the effectiveness of widespread recycling programs. We investigate whether an auditory prompt, crafted to improve recycling accuracy by mitigating inattention to sorting guidelines, influences recycling behavior. To identify the most effective message, we used a large language model (LLM) to simulate survey responses from a representative U.S. sample, selecting the prompt predicted to yield the greatest behavioral improvement. Then, in a field experiment across 24 disposal locations over 30 separate days, we deployed motion-activated audio devices playing the selected prompt – "Please recycle right: only place empty and approved items in the recycling bin!" – at half of the locations after a baseline period. While the LLM forecasted a 17-percentage point reduction in contamination, the randomized controlled trial found only a statistically insignificant 5-percentage point decrease (95% CI: -0.11, 0.01). However, bin design, specifically grey bins with smaller openings, was associated with a 9-percentage point decrease (95% CI: -0.15,0.02). Our results underscore the complexities of translating simulation-based behavioral predictions to real-world settings and emphasize the critical importance of field validation for pro-environmental interventions. The study contributes to behavioral economics and sustainability science by testing dual-modality messaging within the framework of Rational Inattention Theory and demonstrates the potential of LLMs as pre-screening tools for behavioral interventions. However, it also reveals that physical design features may be more effective than auditory nudges in guiding pro-environmental behavior.