Last week, Prof. John Domingue and Dr Aisling Third attended a highly influential Dagstuhl Seminar focused on trust, accountability, and self-determination in AI, particularly in systems built on Knowledge Graphs (KGs) (John served as a co-Chair).
Why Knowledge Graphs Matter
With AI increasingly reshaping the world and driving our most powerful digital platforms—from Google and Netflix to Spotify and Facebook—KGs serve as a foundational technology, structuring vast amounts of web data into machine-readable knowledge. However, as the power of AI continues to increase, concerns over transparency, data privacy, and AI accountability continue to grow, prompting discussions on how to align AI broadly with societal benefits and human rights.
Key Seminar Topics
The event brought together global experts from all over Europe, and the US, Japan, Chile, and Brazil, including Jim Hendler, co-author of the seminal 2001 Semantic Web paper with Sir Tim Berners-Lee, and the ensuing discussions explored:
- Computer Using Personal Agents
- Integrating LLMs with Knowledge Graphs
- Developing Knowledge Graph Ecosystems
- Machine-Readable Policy Descriptions (ODRL-based), and
- Evaluating AI-based systems
Big Questions on AI Ethics
Additionally attendees tackled the following critical questions:
- How can AI be made accountable for its decisions?
- What ensures AI systems remain transparent?
- How can users retain autonomy in AI-driven environments?
Outcomes & Impact
The intense week led to the completion of one paper on personal agents and draft versions of several others. The attendees agreed to collaborate over the next weeks to create a manifesto outlining how trust, accountability, and autonomy in AI systems can be enhanced at the role that Knowledge Graphs will play in that.
Looking Ahead
With overwhelmingly positive feedback, many attendees mentioned that it was one of the best seminars they’ve ever attended. As AI regulation tightens worldwide, KMi continues to play a crucial role in shaping responsible and ethical AI practices.
Related links:
