a solution concept from cooperative game theory that assigns each player their exact fair share of the total value created by a coalition
invented by Lloyd Shapley (1953). the only attribution method satisfying all four fairness axioms simultaneously: efficiency (total value is fully distributed), symmetry (equal contributors get equal reward), null player (zero-contribution agents get nothing), additivity (attributions compose linearly across games).
for a coalition $N$ with value function $v$, the Shapley value of player $i$ is:
$$\phi_i(v) = \sum_{S \subseteq N \setminus \{i\}} \frac{|S|!\,(|N|-|S|-1)!}{|N|!} \left[ v(S \cup \{i\}) - v(S) \right]$$
the average marginal contribution of $i$ across all possible orderings in which the coalition forms.
exact computation is $O(n!)$ — intractable at scale. probabilistic shapley attribution approximates via Monte Carlo sampling: compute each transaction's individual $\Delta\mathcal{F}$, sample $k$ random orderings, cluster by affected neighborhood. complexity drops to $O(k \cdot n)$ with $k \ll n$.
in cyber, the coalition is all neurons contributing cyberlinks in an epoch. the value function is the total focus shift $\Delta\pi$. the Shapley value distributes rewards so each neuron earns proportionally to their causal impact on the equilibrium — the only mathematically fair attribution under the four axioms.
Lloyd Shapley won the Nobel Memorial Prize in Economics (2012) for this and matching theory. the value has since become foundational in machine learning (SHAP explanations), mechanism design, and decentralized reward systems.