Explanation Modalities for a Recommender System Optimizing Energy Management in a Solar-Powered Smart Home

Due to the energy crisis, as well as economical and ecological concerns, home energy management became vital. Although photovoltaic systems are increasingly being adopted by homeowners, understanding how energy is managed to optimize their consumption remains a challenge. We enhance an existing Recommender System (RS) developed to provide daily recommendations for scheduling energy-intensive households activities during solar production peaks, to deliver explanations. This study’s aim is to identify the ideal approach to offer these explanations. Four explanation conditions were evaluated, covering diverse modalities: textual and hybrid (text + visual), with textual explanations generated using an Large Language Model (LLM) and provided in two lengths (short, long). An online questionnaire with 27 respondents was conducted to evaluate these conditions across four seasonal recommendations. Results revealed a preference for the hybrid explanations, suggesting the potential this modality might hold.