Explainable AI for Non-Experts: Energy Tariff Forecasting

Hongnan Ma, Kevin McAreavey, Ryan McConville, Weiru Liu

Abstract: Non-expert users are increasingly affected by the decisions of systems that rely on machine learning (ML), yet it is often difficult for these users to understand the predictions of ML models. In this paper, we propose a web-based platform to evaluate explainable AI (XAI) for non-experts in the context of time series forecasting, focusing on energy price predictions as an exemplary use case. The XAI methods we consider include local feature importance and counterfactual explanations. The platform relies on gamification to encourage user engagement. Our research objective is to evaluate the effectiveness of these different approaches from the perspective of non-expert understanding of machine learning models.