Consequence-Testing
Consequence-testing is the process by which an observer learns which interventions produce which outcomes, and thereby shapes the distribution of interventions it will apply in the future. Unlike random exploration, consequence-testing is directed: the observer perturbs the system where it expects to see effects, and the expectation is itself the product of prior consequence-testing. The result is a self-reinforcing loop in which the observer's intervention distribution converges on the regions of the system that are most informative relative to its cost function.
The concept is central to the Observer-Indexed Emergence framework and the critique of Effective Information. The uniform intervention distribution assumed by EI is the distribution of an observer that has never tested consequences — an observer with no history, no cost function, and no preferences. Real observers are always shaped by consequence-testing, and their intervention distributions are therefore always non-uniform.
Consequence-testing is not limited to scientific experimentation. A predator tests which stalking patterns produce captures. A child tests which vocalizations produce responses. A market tests which prices produce transactions. In each case, the intervention distribution is shaped by the cost of error: failed hunts, ignored cries, unsold inventory. The distribution that survives is the one that maximizes informative outcomes per unit cost.
The formalization of consequence-testing would require integrating reinforcement learning, Bayesian inference, and control theory into a framework that describes how intervention distributions evolve under feedback. No such framework exists in full generality, but the components are present in the literature on adaptive control, active learning, and embodied cognition.
Consequence-testing is how the world teaches the observer what to ask. The observer that does not test consequences is not objective — it is ignorant. And ignorance is not a virtue in science; it is a stage to be grown out of.