In the world of machine learning, where entities are often seen as independent, relational learning stands out by recognizing the crucial connections between them. This field, equipped with powerful tools like Graph Neural Networks and Knowledge Graph Embeddings, offers a way to understand complex, interconnected data. However, a significant limitation persists: the “black box” nature of these methods, which obscures their underlying decision-making processes. This is where explainable and interpretable methods become essential. By shedding light on how these models work, we can gain valuable insights into the relationships within the data. Moreover, the inherent graphical structure of relational data provides a unique opportunity to develop eXplainable AI (XAI) methods that leverage this structure for interpretation. Despite its potential, this avenue remains largely unexplored in current XAI approaches. Bridging this gap is crucial. Developing interpretable-by-design models and effective XAI methodologies specifically for relational data and methods will not only enhance trust and understanding, but also unlock the full potential of relational learning across various domains. This involves establishing clear theoretical foundations and definitions for XAI in the context of relational learning, paving the way for more transparent and insightful analyses.

  • Interpretable by design relational methods
  • XAI for Knowledge Graphs
  • XAI for Knowledge Graphs Embeddings
  • XAI for Graph Neural Networks
  • Neuro-symbolic AI in relational domains
  • Hybrid relational models combining XAI with structured and unstructured knowledge
  • Exploiting LLMs in explaining relational methods
  • Probabilistic methods in the relational domain
  • Ethical considerations of relational XAI
  • Applications of relational XAI such as in healthcare, finance, malware detection and governance
  • Beyond explainability: interpretability, transparency, and trustworthiness in Relational Learning
  • Causality as part of relational learning
  • Interactive and Human-in-the-loop Relational Learning
  • Interpretable Relational Learning for Specific Domains, like Biomedicine, Social Network Analysis, Natural Language Processing
  • Scalability and efficiency of interpretable relational methods
  • Combining Relational Learning with Deep Learning in an interpretable way
  • Explainable reasoning over Knowledge Graphs
  • Fairness and bias detection in Relational Learning
  • Theoretical foundations of interpretable relational learning
  • Evaluation metrics and benchmarks for interpretability in relational models