[DSRP Evidence](https://dsrpevidence.org/)

# How do mental models become better aligned with reality over time?

## The research claim

Every account of how understanding improves describes updating. Bayesian revision adjusts credences on evidence. Conjecture and refutation discards what fails a test. Experiential learning cycles through action and reflection. Predictive processing minimizes surprise. Each is about what happens to the information in a model. This one is about what happens to its organization. A model fails because it is arranged differently from the thing it models, and the failure is the signal. Not an obstacle to understanding — the mechanism of it. A plan that does not work is reality reporting a structural mismatch, and the report contains the correction. The loop has two directions and needs both. Fitback runs outward: the model is put to the test against what the world will allow. Feedback runs inward: the world corrects the model. Block either one and learning stops. A model never tested cannot be corrected, and a model that cannot be corrected is only defended. The condition that keeps both open is corrigibility — the capacity of a system to be changed by what it encounters. This is a structural requirement, not a virtue. Humility and curiosity are what corrigibility looks like from the inside, and a system without it does not approach reality no matter how much information it takes in. Over time, with both directions open, the ratio of model to reality approaches one. The limit is one; the space between 1/1 and 1 is zero. But no finite moment arrives there, and that is not a shortcoming — total collapse of model into world would leave nothing left to learn. The claim is about the direction of travel. Two things follow that are not obvious. Misunderstanding is not evidence that the loop has failed — it is the realization of misunderstanding that starts it. And effort is not enough. The loop needs structure that can carry feedback: boundaries, reciprocity, coherent roles, the capacity to take the other's view. Below that threshold, trying harder does nothing at all. The name for running this loop deliberately is love. Not a feeling — a behavior. Loving reality more than you love your mental model (your preferred story), again and again.

## What would confirm it

Over time, and where both directions of the loop are open, the gap between a model and the thing it models closes. Read in symbols: as time goes on without limit (lim as t→ ∞), the ratio of the mental model at that time (𝕄ₜ) to reality (ℝ) equals one. A ratio of one means the model is organized the way the thing is organized. The limit is stated as equality because the difference between 1/1 and 1 is zero, though no finite moment arrives there. The expanded form says where a gap comes from. The model is the information it holds times the organization applied to it (I_𝕄 · O_𝕄); reality is the information in it times the organization inherent to it (I_ℝ · O_ℝ). A model can fall short by missing information or by being organized differently from the thing. Those are separate failures and they have separate repairs.

## What would refute it

The gap does not close. Models are tested against the world, the world answers, the answers are received — and on average the mismatch stays the same or grows. Read in symbols: the expected value (E) of the rate of change over time (d/dt) of the distance between the model and reality (d(𝕄ₜ, ℝ)) is greater than or equal to zero. Distance not decreasing is the whole of it. The conditions attached matter: this is the null given fitback that actually tests, feedback that actually returns, and structure able to carry both. A model that fails to improve while sealed off from correction says nothing about the loop, because the loop was never running.

## The evidence, and why this status

The convergence claim is measured in people at three time scales, by three different methods, and it points the same way each time. Over years, a panel study found that where one partner's view failed to match the other's lived reality, corrective behaviour followed later. Over days, a diary study found that people who tried harder to understand their partner reported more closeness the next day. Over minutes, couples whose conversations carried more structural alignment had higher relationship quality. Years, days and minutes fail in completely different ways, and none of them relies on self-report about the relationship — the criterion is another person's life, which the model does not control. One study forbids something rather than finding it. Agents ran through repeated cycles of sampling reality, updating and acting again. Below a threshold of organizational integrity the loop stalled however much effort was applied; above it, better than 95% of runs converged. Effort without structure does not close a gap, and the threshold says where. Underneath all of it sits the machinery the loop runs on, established separately: understanding rises as the product of information and organization, and collapses when either goes to zero. Two entries here do not support the claim. One is external work that narrows it, tightening what can be said about the limit. The other is the loop failing at scale — people rating solutions drawn from an independently established web of causes, and looking for the single one that works rather than judging them all worth doing. The rest state the theory, the collective version of the same loop, and the evolutionary argument for why the structure is there at all.

## Supporting evidence (10 publications)

- [Cabrera D, Cabrera L (2026). Love Is a Behavior, Not a Feeling: A Testable Theory of Love as a Reality Alignment Process. Journal of Systems Thinking 6(1):1-31.](https://dsrpevidence.org/paper/602)

- [Cabrera D, Cabrera L (2025). The structure of thought and reality: empirical evidence for the three universal laws linking mind, matter, and meaning. Journal of Systems Thinking.](https://dsrpevidence.org/paper/504)

- [Cabrera D, Cabrera L (2025). From DSRP Theory to O-Theory: The Science of Organization. Journal of Systems Thinking 5(1):1-8.](https://dsrpevidence.org/paper/473)

- [Klein M (2026). The Asymptotic Boundary of O-Theory: Why 1/1 ≠ 1. Journal of Systems Thinking 6(1).](https://dsrpevidence.org/paper/657)

- [Steinhall N, McPettit R, Bond J, Parks M, Khan M, Sharfarz D, Cabrera L, Cabrera D (2024). Wicked Solutions for Wicked Problems. Journal of Systems Thinking 4(1):1-69.](https://dsrpevidence.org/paper/454)

- [Cabrera D, Cabrera L (2025). The Science of Getting on the Same Page: The VMCL Field Equation Behind Adaptive Culture. Journal of Systems Thinking 5(1):1-33.](https://dsrpevidence.org/paper/503)

- [Cabrera D (2006). Systems Thinking (Doctoral Dissertation). Cornell University.](https://dsrpevidence.org/paper/86)

- [Cabrera D, Cabrera L (2026). Scale-Free Structure: Is Knowledge Just the Recurring Emergence of a Universal Structural Grammar?. Journal of Systems Thinking 6(1):1-40.](https://dsrpevidence.org/paper/633)

- [Cabrera & Cabrera (2025). Oneness. Odyssean Press.](https://dsrpevidence.org/paper/480)

- [Cabrera, Derek (2026). The Patches Are the Evidence: Convergence on DSRP-484 Across System Dynamics, Network Theory, and UML. Journal of Systems Thinking.](https://dsrpevidence.org/paper/691)

[All research questions](https://dsrpevidence.org/claims)