Baselines, Counterfactuals & Prediction Error

Primes about interpreting an outcome against a reference point rather than in isolation: counterfactual subtraction and baseline deviation estimate effects by comparing to what would otherwise have happened, reward prediction error and counterfactual proximity weighting use the signed gap as a teaching signal, and residual analysis mines the leftover for structure.

7 primes in this family — primes that sit near one another in abstraction space (k-means over structural-signature embeddings). Each is shown with its short description.

• Anticipatory Neutralization — Forward-looking agents pre-adjust to offset an anticipated intervention's intended effect.
• Baseline Deviation — An observation is interpreted against a declared reference and flagged as departing from it, producing deviation as a first-class fact.
• Counterfactual Proximity Weighting — An internal signal is graded not by the outcome alone but by its distance to a nearby counterfactual outcome of different value, so near misses carry near-reward signals.
• Counterfactual Subtraction — An effect is estimated by subtracting a constructed baseline representing what would have obtained absent the intervention, so the inference rests on the baseline's credibility, not the arithmetic.
• Outcome-Defined Adequacy — Adequacy is specified by whether a defined outcome is produced in context, leaving the form open to any realization that achieves it.
• Residual Analysis — Subtract the best explanation and study the leftover as the next site of structure, not as noise.
• Reward Prediction Error — The signed difference between received and expected reward or value becomes a teaching signal for updating value estimates, policies, and reward-linked behavior.