FIRE: A Modular Framework for Prototyping and Evaluating Interactive Recommender Explanations

Explanations in recommender systems are important for enhancing transparency, trust, and system understanding. However, as research interest in interactive and chat-based explanations grows, the absence of standardized, reusable interface components limits their systematic evaluation. We introduce the Framework for Interactive Recommender Explanations (FIRE) to bridge this gap. FIRE provides an open-source library of modular explanation styles with varying levels of interactivity — spanning from static lists and draggable charts to conversational agents. To test these modalities in online user studies, FIRE integrates a supporting serverless experiment engine equipped with dynamic routing, crowdsourcing integration, and complete interaction logging. Currently configured for group recommenders but built for broader applicability, FIRE helps to standardize the evaluation of the explanation layer, thereby driving reproducible experimental design.