Best Design Practices for Hybrid Neuro-Symbolic Systems

During our ORKG research group retreat in Goslar, Germany, held on May 15–16, 2025, I had the privilege to present part of my PhD work entitled: "Best Design Practices for Hybrid Neuro-Symbolic Systems" - exploring how we can make AI both powerful and human-friendly. What are Neuro-Symbolic Systems? These systems represent a seamless integration between: Neural networks: Fast, intuitive learning (analogous to System 1 thinking) Symbolic reasoning: Logical, rule-based processing (analogous to System 2 thinking) This mirrors how human cognition operates through both quick instincts and deliberate reasoning. Why does this matter for HCI? Traditional AI often functions as a "black box" - powerful but opaque. Neuro-symbolic systems offer transparency and explainability, but only when properly designed with human-computer interaction principles in mind. Key Design Guidelines Presented: Define Appropriate Scope: Narrow-focused features outperform broad, general-purpose ones in usability Design for Mental Models: Help users develop accurate expectations of AI capabilities and limitations Improve Discoverability: Make AI features easily findable and recognizable through familiar UI patterns Ensure Transparency & Trust: Display confidence levels, cite sources, and communicate uncertainty appropriately Trust, but Verify: Enable human oversight and validation, especially for critical decisions Design for Co-Creation: Position AI as a collaborative partner rather than a replacement for human creativity Research Contribution: This work demonstrates how tracking user edits can serve as a quality metric for AI-generated content - fewer required edits indicate higher AI accuracy and increased user trust. The presentation includes real-world examples from our data trustworthiness reporting interface and research problem analysis system. Significance: This research contributes to the growing field of human-centered AI design, emphasizing that the future of artificial intelligence lies not merely in increasing computational power, but in creating systems that are truly human-centered and collaborative.