The Field Guide to Black Box Prediction The Fifteen Rules

The Field Guide to Black-Box Prediction presents a set of fifteen operational heuristics for predicting the behavior and failure modes of opaque systems when internal structure is unavailable, unreliable, or irrelevant. Rather than offering a theory of system internals, the guide focuses on observable behavior under constraint, pressure, error, and correction. The rules function as diagnostic tools for identifying hidden incentives, unmodeled constraints, unstable decision boundaries, and suppressed feedback loops in systems ranging from artificial intelligence and algorithmic decision processes to institutions and complex organizations. The guide is intended as an applied companion to recursive epistemic modeling approaches, translating abstract meta-epistemic principles into practical checks that can be used by observers without privileged access or interpretive authority. Emphasis is placed on prediction over post hoc explanation, with particular attention to conditions under which explanatory coherence masks growing divergence from reality. By framing prediction as the primary epistemic virtue in black-box contexts, the field guide offers a compact, transferable method for anticipating systemic failure before it becomes visible or irreversible.