Every institution has an ideology---a set of frames that filter perception and guide behaviour. But the ideology isn't what's written in the values statement. It's what's enacted in daily practice. The stated values are aspirational; the enacted values are operational. And the brain, which maps regularities, learns the operational ones.

This is why institutional hypocrisy is so corrosive. When espoused and enacted values diverge, the system doesn't split the difference---it maps the enacted values and discounts the espoused ones. Worse, the divergence itself becomes part of the ideology: people learn that values statements are performative, that the institution doesn't mean what it says, and that the real game is reading the unwritten rules.

So what can you do? Close the gap from the enacted side, not the espoused side. Adding more values training when the daily experience contradicts the values doesn't help---it reinforces the lesson that values are talk, not action. Instead, change what actually gets rewarded and punished. Promote people who embody the values. Correct violations visibly and consistently. The ideology updates through practice, not through policy.

The input-output machine learns from what the environment provides. If the environment provides rewards for corner-cutting, that's what gets learned. If it provides consequences for violations, that's what gets learned. The rules on the wall are inputs too---but they're weak inputs, easily overridden by the much louder inputs of daily experience.

Think about monitoring density: how often do supervisors actually observe what their people do? In organisations with low monitoring density, the gap between stated and enacted values can grow enormous, because there's no environmental input reinforcing the stated rules. In organisations with high monitoring density, the gap shrinks---not because people become more virtuous, but because the environmental inputs are more consistent with the stated values.

But monitoring density has a dangerous failure mode: it tightens adherence to whatever norms the monitoring network enforces. When institutional-level oversight is weak but unit-level monitoring is strong, the mechanism can tighten adherence to deviant norms rather than institutional standards. The unit's enacted values become tightly enforced by peer pressure, while the organisation's stated values become irrelevant. This is how systematic violations can persist within a tightly disciplined group---the discipline is real, but it's disciplined adherence to the wrong norms.

So what can you do? Design the environment so the stated values have environmental inputs backing them up. This doesn't mean surveillance---it means presence, visibility, and consistency. Leaders being present, seeing what happens, and responding to it is the single most powerful environmental input for aligning enacted and espoused values. The social map updates from what it observes, and it observes constantly. And ensure monitoring operates at the right level. Unit-level monitoring alone is insufficient---if it can tighten around deviant norms, it will. External oversight, cross-unit observers, independent reporting chains, and rotation of personnel all ensure that institutional norms are what tightens, not local subculture.

The prediction engine builds its model from experience. If experience says "violations get overlooked," the system predicts that violations will continue to get overlooked, and it adjusts behaviour accordingly. If experience says "this leader means what she says," the system predicts consistency and adjusts behaviour to match. The prediction is built from every data point the brain has accumulated, and the stated values are a tiny fraction of that data.

This is why leadership transitions are so powerful---or so dangerous. A new leader changes the prediction. But the old predictions are deeply encoded, and a few instances of new behaviour won't override years of the old pattern. The new behaviour has to be consistent enough, and sustained enough, for the prediction engine to update.

So what can you do? Be relentlessly consistent, especially in the first months of any change. The prediction engine is watching, and it's looking for evidence that the new rules are real. Every inconsistency resets the counter. Every exception teaches the system that the new rules are as performative as the old ones. Consistency over time is the only signal strong enough to rewrite a deeply encoded prediction.