Parenting Horizons

Follows The Hypothesis Graph.

My child approaches a crosswalk. I don’t hold their hand. They check the signal, check for cars, cross. I’ve tested this before — graduated withdrawal over months, watching whether the signal was internalized. It was. The perturbation confirmed it.

The behavioral interventionist watching classified it as misbehavior.

Same event, two rational readings, no shared language to resolve the disagreement.

The window determines the bin

Research on children typically lasts 2–5 years. That’s considered long-term. Parenting spans two decades. A parenting intervention that looks convergent at year 3 might be oscillatory at year 10. The study window determines which bin you see, and funding cycles set the study window.

This is the p-value problem applied to the longest system most people will ever operate. Compress a 20-year trajectory into a 3-year snapshot, and the temporal structure is gone. The intervention “works” if the child’s behavior improves within the study window. Whether it created a dependency, an oscillation, or a fragility that only surfaces when the scaffolding is removed at 18 — that’s outside the window. Nobody is watching.

Three bins the interventionist can see

Behavioral interventionists are trained to keep kids safe. Their classification is binary: safe or unsafe. The hypothesis graph offers finer resolution.

Convergent stress. Challenge meets the child at the edge of capability. The child adapts, builds the skill, and the stress resolves. This is learning. A child who struggles with a puzzle, gets frustrated, and solves it has run a convergent perturbation. The stress was the signal that the current strategy was insufficient.

Divergent stress. Overwhelm past the child’s capacity. Shutdown, cascade, damage. A child screamed at for something they can’t control. The stress doesn’t resolve because the child has no strategy available. This is what interventionists are trained to prevent, and they’re right to prevent it.

Oscillatory stress. The permissiveness cycle. Parent is too permissive, child pushes boundaries, parent snaps to authoritarian, feels guilt, returns to permissive. Two subsystems fighting, period set by the parent’s tolerance threshold. Any 2-year study that starts in the permissive phase and ends after the first correction will call it stable. The oscillation only shows up if you watch long enough to see the second cycle.

The interventionist’s training distinguishes divergent from safe. It doesn’t distinguish divergent from convergent — because that classification requires the trajectory, and their window is a shift.

Controlled perturbation

I withdraw attention at regular intervals. Not as punishment — as a test. Does the child’s behavior converge back to self-regulation, oscillate between testing and compliance, or diverge into distress?

This is the mechanic tapping the alternator. Bounded perturbation, long enough to classify, short enough to be safe. The response tells me whether the capability is internalized or whether I’m the compensating mechanism.

The interventionist sees withdrawal of attention and classifies it as negligence. She’s applying the divergence-prevention rule to a convergence test, because her framework has no category for “the parent already tested this and it converged.” Her classification scheme doesn’t have kill conditions. It has a checklist.

The incentive horizon

The disagreement generalizes. Whenever the evaluation window is shorter than the system’s dominant timescale, the optimizer produces convergence within the window at the cost of oscillation or divergence outside it.

Behavioral interventionists are evaluated per shift. CEOs are evaluated quarterly. Drug trials run 9 months (Vioxx needed 18 to reveal the cardiovascular trajectory). A/B tests run weeks on systems with monthly feedback loops.

Munger had the folk version: “Show me the incentive and I’ll show you the outcome.” The trajectory is determined by the incentive structure’s timescale, not the system’s. He never formalized it either — another practitioner’s heuristic refined over decades, no math underneath.

When Musk secured his payout with no guarantee of a future one, he stopped optimizing for Tesla and started building rockets. Remove the short-window incentive, observe the trajectory. Divergence — the compensating mechanism was the compensation package. The stock price was convergent because the payout kept his attention. Remove it, and you see what the system actually wants to do.

The formalism gap

A doctor can say: “I tested for X, the result was Y, so I ruled out Z.” The vocabulary exists. The institution accepts trajectory-based evidence because medicine formalized differential diagnosis centuries ago.

A parent cannot say: “I tested the child’s crosswalk internalization at three prior intersections with graduated withdrawal and observed convergence, so hand-holding is no longer indicated.” The sentence doesn’t exist in the interventionist’s training. So evidence that is real, sequential, and diagnostic is illegible to the institution evaluating the parent.

That’s what the hypothesis graph’s open question about parenting actually means in practice. Not that nobody has modeled the dynamics — they have. But the epistemology of parenting as a learned practice remains unformalized. There is no shared protocol for a parent and an interventionist to resolve a disagreement about whether a perturbation is convergent or divergent. Both are rational. Neither can convince the other. The formalism would be the shared language.

Horizon

An overprotective parenting style produces convergence on every observable metric for 15 years. The child is safe, regulated, compliant. The divergence appears when the scaffolding is removed — the child leaves home and can’t self-regulate because the parent was the regulating mechanism. What looked like convergence was a system that was divergent the whole time, masked by the compensating mechanism.

The trajectory framework predicts this: “Converging null: there was an early signal that decayed. The system compensated. Edge: test the compensating mechanism.” Except in parenting, you can’t run that test without removing the mechanism, and removing it is the child leaving home.

Unless you run small, bounded, reversible versions of that test along the way. That’s what controlled perturbation is. The parent who periodically withdraws scaffolding and watches the response is sampling the trajectory that the 18-year removal will reveal in full. Each sample tells you whether you’re building self-regulation or dependency.

The interventionist calls it negligence. I call it the only way to see the trajectory before it’s too late to change it. But even recognizing the gap is the start of a convergent conversation.