where 1\gamma\ge1 tunes concentration (higher \gamma makes the highest score dominate).
3) Algorithm: ensemble simulation + actor attribution
Inputs: calibrated model parameters, initial states P0,T0,E0,H0P_0,T_0,E_0,H_0, noise model, candidate actor scores time series si(t)s_i(t), emergence threshold HcH_c, ensemble size NN.
Pseudo-code:
for k = 1..N:
  simulate X_t^(k) = {P,T,E,H,R} forward to horizon T_max with SDE solver
  record _k = first t where H^(k)(t) > H_c (or _k = if not crossed)
  if _k < :
    record actor selection at _k: sample actor i with probability proportional to s_i(_k)^
end
# Compute emerg probability curve