Decision Support Based on Large Language Models: Ontology-Based Architecture and Generalized Scenario

Le résumé fourni par la source

The paper presents initial results on models and methods of LLM-assisted augmented intelligence in context-aware decision support. It investigates a range of LLM-based decision support scenarios and reveals that they face common challenges, including problem clarification, user preference consideration, domain knowledge usage, generating alternatives and making decisions. Based on these findings, a set of requirements for LLM-based decision support is specified, an architecture for an LLM-based decision support system is proposed, and a generalized LLM-based decision support scenario is developed. The architecture combines agent-based interaction, conceptual modeling, and external solvers to support flexible and explainable decision-making. It enables the dynamic adaptation of recommended decisions to user preferences and evolving contexts. The generalized scenario incorporates the common challenges into a unified workflow. It integrates ontology-driven knowledge representation, conversational problem modeling, and problem solving by computational components. Within the framework of this scenario, the specific scenarios of LLM-assisted ontology development and decision support are discussed in detail. The contribution provides a foundation for developing LLM-based decision support systems enabling effective and informed decision-making.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.