We argue that integrating Knowledge Graphs (KGs) with Generative AI (GenAI) models provides context-aware, adaptive explanations in dynamic environments. By leveraging structured knowledge from KGs, GenAI systems can ofer explanations that adjust to user needs and real-time contexts, enhancing transparency and user trust. Generative AI (GenAI) models have achieved impressive results across various domains, from natural language processing to image generation [1]. However, one of their critical shortcomings is the lack of transparency in their decision-making processes [2]. The 'black-box' nature of GenAI systems becomes particularly critical in dynamic environments where context is vital, such as autonomous systems, personalized recommendation engines, and adaptive decision-making tools [3].
KGs are structured repositories of interconnected facts and relationships, which researchers and practitioners can leverage to provide explanations grounded in factual knowledge [6]. For example, a KG in the healthcare domain might contain relationships between diseases, symptoms, and treatments, enabling an AI model to generate detailed explanations for a medical diagnosis [9].
CEUR
ceur-ws.org
GenAI models can produce multi-layered explanations by integrating KGs into their architecture [6]. Consider a recommendation system in an e-commerce setting. The system might first explain why a customer received a product recommendation, such as ‘based on browsing history’. With the help of a KG, the user could then explore deeper layers of the system’s reasoning, such as specific item similarities or product ratings, thus adjusting the depth of the explanation to the user’s needs [10].
Personalised recommendation engines, like the ones being utilised in retail, e-commerce and streaming services, KGs can enhance explainability by drawing on user preferences, historical data, and item relationships [11]. For instance, if a movie recommendation is generated, the system could explain it by showing connections in the KG, such as genre preferences, actors or similar movies the user has watched, creating a tailored, context-sensitive explanation.
Autonomous vehicles are an example of a dynamic environment where context-aware explanations are critical. When making decisions, such as selecting a route or dealing with sudden obstacles, the system must explain its choices in real time [12]. A KG-based system could provide high-level explanations like ‘avoiding trafic congestion’ while ofering deeper insights such as weather conditions, real-time trafic data, and road safety statistics, all drawn from the KG.
Despite the potential benefits, several challenges must be addressed before deploying KGs in real-world applications. First, constructing and maintaining KGs for specific domains is resource-intensive, which might hinder their development in SMEs. Second, the computational overhead of querying largescale KGs in real-time environments remains a constraint for most public and private organisations. Consequently, future studies should focus on developing automated or semi-automated methods for KG creation and maintenance as well as optimising the scalability of KG-query processes.
Future work needs to be done to refine the mechanisms of adaptative explanation. For instance, understanding the best ways to balance explanation depth and usability to maximise the impact of KG-based applications, or exploring how to handle conflicting explanations in large and multi-source KGs solutions.
We argue that Knowledge Graphs are necessary to enhance explainability in Generative AI models. By providing context-aware, dynamic explanations, we can significantly improve the transparency and user trust in AI systems, particularly in dynamic environments like autonomous systems and personalised recommendations. While challenges remain, integrating KGs into explainable AI is a promising direction for future research with substantial real-world applications.
During the preparation of this work, the author(s) used GPT-4 and Grammarly for Grammar and spelling checks. After using these tool(s)/service(s), the author(s) reviewed and edited the content as needed and take(s) full responsibility for the publication’s content. [2] P. P. Angelov, E. A. Soares, R. Jiang, N. I. Arnold, P. M. Atkinson, Explainable artificial intelligence: an analytical review, Wiley Interdisciplinary Reviews: Data Mining and Knowledge Discovery 11 (2021) e1424. [3] D. Minh, H. X. Wang, Y. F. Li, T. N. Nguyen, Explainable artificial intelligence: a comprehensive review, Artificial Intelligence Review (2022). [4] M. Gaur, A. Desai, K. Faldu, A. Sheth, Explainable ai using knowledge graphs, in: ACM CoDS
COMAD Conference, 2020. [5] M. Gaur, K. Faldu, A. Sheth, Semantics of the black-box: Can knowledge graphs help make deep learning systems more interpretable and explainable?, IEEE Internet Computing 25 (2021) 51–59. [6] E. Rajabi, K. Etminani, Knowledge-graph-based explainable ai: A systematic review, Journal of
Information Science 50 (2024) 1019–1029. [7] J. Schneider, Explainable generative ai (genxai): a survey, conceptualization, and research agenda,
Artificial Intelligence Review 57 (2024) 289. [8] B. Y. Lim, A. K. Dey, D. Avrahami, Why and why not explanations improve the intelligibility of context-aware intelligent systems, in: Proceedings of the SIGCHI conference on human factors in computing systems, 2009, pp. 2119–2128. [9] F. Teng, W. Yang, L. Chen, L. Huang, Q. Xu, Explainable prediction of medical codes with knowledge graphs, Frontiers in bioengineering and biotechnology 8 (2020) 867. [10] L. Xie, Z. Hu, X. Cai, W. Zhang, J. Chen, Explainable recommendation based on knowledge graph and multi-objective optimization, Complex & Intelligent Systems 7 (2021) 1241–1252. [11] M. H. Syed, T. Q. B. Huy, S.-T. Chung, Context-aware explainable recommendation based on domain knowledge graph, Big Data and Cognitive Computing 6 (2022) 11. [12] S. N. N. Htun, K. Fukuda, Integrating knowledge graphs into autonomous vehicle technologies: A survey of current state and future directions, Information 15 (2024) 645.