Update on work on chunk based cognitive agents

I’ve been working on adapting ideas from John R. Anderson’s work on ACT-R, as a basis for web-based cognitive agents. This provides a framework for data and rules based upon chunks - a term from psychology for a collection of things that are easier to remember as a group. Chunks embrace both RDF triples and Property Graphs. For more information on chunks and rules see:

 https://www.w3.org/Data/demos/chunks/chunks.html <https://www.w3.org/Data/demos/chunks/chunks.html>
 https://www.w3.org/Data/events/tpac2019/digital-transformation.pdf <https://www.w3.org/Data/events/tpac2019/digital-transformation.pdf>

This may be the wrong list to ask for these, but in general, collecting challenges for researchers to gauge progress against would be really helpful for driving progress, and help to evaluate which approaches are more effective.

I am interested in use cases and datasets suitable for use in work on, e.g.

Unsupervised learning of taxonomies and ontologies from noisy data
Reinforcement learning of skills in simulated environments
Causal reasoning, for planning actions or explaining faults

Note that I am currently working on a demo featuring autonomous driving as a simulated environment that will allow me to explore how different cognitive tasks can coexist on the same rule engine and work together in a timely way. In principle, this could be extended to support reinforcement learning scenarios.

Dave Raggett <dsr@w3.org> http://www.w3.org/People/Raggett
W3C Data Activity Lead & W3C champion for the Web of things