Software Diagnostics and Observability Institute

Representation access and functional dissociation in pattern-oriented diagnostics

PDF: https://www.dumpanalysis.org/files/Neurolinguistic_and_Neurosymbolic_App...

Finding data in a memory dump does not mean the application can use it. A document may remain intact while its lookup path is broken. An obsolete object may preserve recognizable contents after its useful lifetime. Such distinctions connect the structures we observe to the operations we need to explain.

In this article, we explore how neurolinguistics and cognitive neuropsychology can inform pattern-oriented software diagnostics through representation and access, functional dissociation, disconnection, and inappropriate persistence. We then connect these investigative distinctions to a proposed neurosymbolic architecture in which neural components suggest patterns and further observations, while symbolic components evaluate explicit conditions against debugger evidence.

The design builds on Analysis Pattern Networks, Trace Ontology, Trace World, our PrologWinDbg extension, and our published book Prolog for Pattern-Oriented Software Diagnostics. Worked scenarios show how evidence, coverage, and assumptions constrain a diagnostic conclusion. The article develops a system design, examines its proposed novelty, and identifies further directions for implementation and evaluation. Throughout, we distinguish existing capabilities from proposed extensions and supported structural findings from causal hypotheses.