Symbiotic Knowledge Graphs: A Vision for Semantic Brain-Computer Interfaces

Post-WWII: bottleneck shifts from scientific knowledge production → consumption

Vannevar Bush, Manhattan Project, NSF Photo: Office for Emergency Management, 1940–44 (Library of Congress, public domain)

Memex illustration by Alfred D. Crimi, LIFE, 10 Sept. 1945

• Bush argued the need for the Memex

  Device to store all forms of data, and links between them
  As We May Think, 1945 essay

Need for cooperative interaction between humans and machines

Joseph Carl Robnett Licklider, ARPA Photo: MIT Museum

• Computers should not only automate calculations

  Humans: creativity and intuition
  Computers: precise calculations and information
  Man-Computer Symbiosis, 1960 essay
  
  

Licklider, IRE Transactions on Human Factors in Electronics, 1960

Humans solving complex problems better by directly augmenting human cognition

• Do not only focus on making computers more intelligent

  Humans should also become better at solving complex problems
  → computer mouse, windowed interfaces, collaborative editing, …
  Augmenting Human Intellect: A Conceptual Framework, 1962 report
  
  

Douglas Engelbart, Augmentation Research Center Still from the 1968 SRI demo

Engelbart, Augmenting Human Intellect, SRI report, 1962

The Semantic Web → intelligent agents

Tim Berners-Lee, Jim Hendler, Ora Lassila Photos: Knowledge Graph Conference (KGC), 2025

Illustration © 2001 Scientific American (The Semantic Web, Berners-Lee, Hendler & Lassila)

• Machine-understandable layer on the Web

  Intelligence is external to humans (Licklider)
  The Semantic Web, 2001
  
  

• → Let's revisit Engelbart's vision!

  Not only increase artificial intelligence
  but also increase human intelligence.

Why does this matter now?

Google effect: we offload thinking to search engines and AI, causing our memory and skills to degrade (Sparrow et al., 2011).

Integrating Brain-based Knowledge Graphs with external Knowledge Graphs

Using a (hypothetical) Semantic Brain-Computer Interface

Semantic Web stack provides important building blocks

• Align internal concepts of different people

  IRIs and ontologies

• Integration across brain data and Web data

  SPARQL federation

• Shared reasoning across humans and machines

  Reasoning, proof languages, neuro-symbolic AI

Opportunities using current BCI technology

BCIs can encode and decode sensory information

• 

  Human-driven external processes and data generation

  Requires modeling human senses and actuators (SSN?)
  e.g. interoperable mechanical operations using human actions
• 

  Human-driven semantic data exploration

  Requires mapping link semantics to human actuators
  e.g. Web or Knowledge Graph navigation

BCI technology has major limitations

• Limited understanding of neural encoding

  Abstract concepts and complex knowledge relations

• Limited ability to integrate externally stored memories

  Prosthesis techniques exist for restoring or enhancing memory

• Limited information bandwidth

  May be necessary to exchange large chunks of knowledge

BCIs are rapidly evolving due to increasing commercial interest

Source: byFounders, Brain-Computer Interfaces: Exploring the Last Frontier, 2025

Mapping biological semantic memory to and from symbolic knowledge

• Memories are combinations of personal multimodal experiences

  e.g. visual features, taste, smell, …

• Concepts across people are not necessarily stored the same

  Personal experiences differ

• Need for mappings to align biological memories across people

  Brain to RDF Mapping Language (B2RML)?

Mapping biological reasoning to and from (sub)symbolic reasoning

Augment human reasoning with external reasoners

• (Sub)symbolic AI enables reasoning over digital knowledge

  → AI reasoning over brain knowledge
  → Human reasoning over external knowledge

• Need for exchanging reasoning language and contexts

  Exchange LLM context windows, symbolic reasoning, proof languages, …

Modeling ownership and access control for biological memories

• Memories are highly personal

  → Major ethical concerns

• Need for strong access control and consent mechanisms

  Are current authorization mechanisms sufficient?
  E.g. RBAC, ACL, ODRL, …

Conclusions

• Human Intelligence and Artificial Intelligence are at a crossing point

  What happens if (significantly) AI exceeds HI?
  → Let's augment Human Intelligence

• Semantic Web stack provides foundation for intelligent agents

  It can also provide the basis for SBCIs
  → Concrete research opportunities already exist today

• Strongly depends on BCI breakthroughs

  Future opportunities may not be far off
  → Let's team up with cognitive neuroscience and BCI researchers