Stakeholder Suite: A Unified AI Framework for Mapping Actors, Topics and Arguments in Public Debates

TL;DR

The paper tackles the challenge of understanding public debates around energy infrastructure by mapping stakeholders, topics, and arguments in a transparent, end-to-end framework. Stakeholder Suite integrates LLMs and Retrieval-Augmented Generation to produce a unified representation and enable scalable extraction of arguments with stance while grounding outputs in sources. The methodology encompasses data collection, actor detection, topic thesaurus construction, argument mining across global/actor/topic dimensions, and rich visualizations, validated on multiple renewable-energy projects. Evaluations show robust retrieval precision and stance accuracy, with user pilots confirming operational utility, though the LLM judge can be conservative and GDPR considerations are addressed. The work contributes a production-ready framework bridging gaps between commercial and academic approaches to controversy analysis.

Abstract

Public debates surrounding infrastructure and energy projects involve complex networks of stakeholders, arguments, and evolving narratives. Understanding these dynamics is crucial for anticipating controversies and informing engagement strategies, yet existing tools in media intelligence largely rely on descriptive analytics with limited transparency. This paper presents Stakeholder Suite, a framework deployed in operational contexts for mapping actors, topics, and arguments within public debates. The system combines actor detection, topic modeling, argument extraction and stance classification in a unified pipeline. Tested on multiple energy infrastructure projects as a case study, the approach delivers fine-grained, source-grounded insights while remaining adaptable to diverse domains. The framework achieves strong retrieval precision and stance accuracy, producing arguments judged relevant in 75% of pilot use cases. Beyond quantitative metrics, the tool has proven effective for operational use: helping project teams visualize networks of influence, identify emerging controversies, and support evidence-based decision-making.

Figures (5)

• Figure 1: Blueprint of the Stakeholder Suite framework with an actor argument example: (1) Construction of a database containing stakeholders and documents related to the public debate; (2) Parsing of documents into paragraphs and detection of actor mentions and interventions within the text; (3) Linking of paragraphs to topics from the thesaurus using semantic similarity; (4) Generation of arguments through a RAG pipeline along three analytical axes (global, topic-specific, actor-specific) together with automatic stance classification for each argument.
• Figure 2: Argument page where we can access all the arguments related to the project and filter by stance, type (actor, topic, global) and by date.
• Figure 3: A zoom on the argument page with actor type filter selected for the Federal Government.
• Figure 4: Actor view page where we can find the actor information, its arguments and the other actors he is connected with.
• Figure 5: Mapping page where we can see the connections of actors and their importance in the debate.