TRACE: A Conversational Framework for Sustainable Tourism Recommendation with Agentic Counterfactual Explanations

A team of researchers led by Ashmi Banerjee has introduced TRACE (Tourism Recommendation with Agentic Counterfactual Explanations), a novel conversational framework designed to combat the environmental downsides of traditional travel algorithms. Unlike conventional systems that optimize purely for user relevance—often reinforcing overcrowded, carbon-intensive destinations—TRACE employs a multi-agent, LLM-based architecture to actively promote sustainable tourism. Built on Google's Agent Development Kit, the framework uses specialized agents to elicit latent sustainability preferences, construct structured user personas, and generate recommendations that balance personal relevance with environmental impact.

A core innovation of TRACE is its use of 'agentic counterfactual explanations' and LLM-driven clarifying questions. Instead of simply presenting a list of options, the system surfaces greener alternatives and prompts users with reflective questions to refine intent, fostering sustainable decision-making without coercion. The researchers have made the project fully reproducible, providing complete code, a Docker setup, prompts, and a demo video. User studies and semantic alignment analyses presented for SIGIR '26 show that TRACE effectively supports eco-friendly choices while maintaining the interactive responsiveness and quality users expect from travel recommenders.