The Research Arc
A chronological view of the questions, experiments, failures, and results that helped shape Computational Meaning Dynamics research. The investigations progress from making meaning measurable, to modeling its dynamics, to testing and shaping semantic behavior in agentic, consequential systems, and human-AI teaming. Across that progression, the work can be understood through three broad shifts in how meaning is observed, modeled, and acted upon.
Foundations
The early work reinterprets machine learning distribution shifts as shifts in the underlying data's semantic meaning. It separates semantic drift from rate effects and shows that semantic and grounding measures reveal changes that conventional similarity metrics miss.
Research Question: How can meaning itself become measurable?
Semantic dynamics
The research shifts from asking whether meaning moved to asking how it moved. Recursive generation, grounding, attractors, transition behavior, and operator-theoretic ideas turn semantic change into a structured dynamical process.
Research Question: If meaning moves, what structures describe it?
Agentic systems and shaping
This phase investigates whether the same observables survive in multi-step, multi-agent, and partially observable systems, extends them into safety-critical and multimodal settings, and asks whether semantic dynamics can be used to shape routing, escalation, and human review.
Research Question: Can semantic dynamics become actionable?