Strategic decision‑making in complex environments requires a trade‑off between local stability and global coordination. Using the game of Go as a bounded empirical testbed, we examine how biological and artificial agents navigate this trade‑off. We formalize a network‑derived strategic tension metric via the spectral radius of spatial dependency matrices:
$$ \text{Tension}=\rho(\mathbf{A}) $$ where $\rho$ denotes the spectral radius of matrix $\mathbf{A}$ that encodes spatial dependencies among board positions.
Across human skill progression, tension systematically decays from novices to masters: beginners engage in frequent uncoordinated skirmishes, while masters stabilize territory and reduce unnecessary conflict. A longitudinal study of professional games from 1980 to 2026, however, reveals a striking historical paradox. After AlphaGo’s superhuman breakthrough in 2016, elite human matches exhibit a marked increase in sustained tension.
We project games onto a two‑dimensional space defined by move distance and captures, positioning modern professionals close to neural engines. Their tension profile diverges from the local contact fighting of novices and from the random noise of unstructured play.
Consequently, strategic tension serves as a macroscopic order parameter that captures decision paradigms rather than raw computational strength. Beyond Go, these findings propose a constructive paradigm for human‑AI interaction: artificial systems can act as cognitive scaffolding, extending biological heuristics and elevating human capacity to navigate higher‑order complexity in professional training and decision‑making domains.
Review