The model you have of them is years out of date

Stage 2 · Months 3 to 12 · Article 33 · Wave 2

---

The Co-Parent in your head is not the Co-Parent of today. Most of the model you're using was built in the late phase of the marriage, when daily friction was at its highest, and hasn't been updated since. You're co-parenting with a model from two or three years ago.

This article covers why the model stopped updating, the four signals that a model is genuinely out of date, the five-question audit that surfaces what's stale, what to do with the new readings, and the asymmetry between updating their model of you and your model of them.

Why the model stopped updating

In an active relationship, your model of the other person updates constantly. Daily contact, observed behaviour, conversations, small moments, all of it feeds into a continuously refreshed picture of who they are now.

Around the time the marriage starts to fail, three things happen to the model.

1. The data became too painful to integrate. In a late marriage, new information about the partner is often unwelcome, it either confirms why the relationship is failing or complicates a decision that's becoming necessary. The model stops updating in real time because real-time updates would require continuous emotional processing.

2. The most extreme behaviours became the dominant data. The model started weighting the worst moments more heavily than the ordinary ones. By the end of the marriage, the worst-case version of them had become the working version. The day-to-day ordinary version was edited out.

3. After separation, the data flow stopped. You no longer have daily contact. The only data points come from messages, handovers, and reports from the children. These are an incomplete sample of who they are. The model freezes at whatever shape it had when the daily contact ended.

By month four or five post-separation, the model in your head is mostly the late-marriage worst-case version, frozen at the moment the daily contact stopped. The Co-Parent has been living, processing, changing for those months. None of that has filtered through.

The four signals that a model is out of date

A few signs that the version you're operating with isn't the actual current person.

Signal 1: You're surprised when they do something you wouldn't have predicted

A reasonable response to a message. A capable handling of a situation with the children. A small kindness. A clear decision. If you find yourself thinking huh, didn't expect that, that's the model failing.

The model predicts based on what it knows. When the prediction misses, the model is older than the data.

Signal 2: They consistently behave better than the model expected

If your model expects bad behaviour and the actual behaviour is repeatedly better than predicted, the model is downward-biased. The late-marriage version of them was probably worse than they are now (or possibly worse than they were at any point, see compression below).

The reverse can also be true. Some Co-Parents are worse post-separation. But the more common pattern in early Stage 2 is that the model overestimates their badness because it was built during the worst phase.

Signal 3: Other people see them differently than you do

If the children, the Co-Parent's friends, mutual contacts, or other observers describe a person you don't recognise, the model may not match what other people are seeing. This isn't conclusive, sometimes others see less than you did. But it's data.

Pay particular attention to what the children report. They have current, continuous data about the Co-Parent. They're a more reliable read than memory is.

Signal 4: The model hasn't changed in months

If your sense of who the Co-Parent is now is identical to your sense of who they were three months ago, the model isn't updating. Even if the actual person hasn't changed, the surrounding context has. The model that doesn't move at all is the model that isn't reading.

The five-question audit

A simple practice to surface what's stale. Write brief answers, in private, to these five questions. Don't show them to anyone.

Question 1: What was the most recent observation that didn't fit your model?

A specific moment in the last few weeks where the Co-Parent did something the model wouldn't have predicted. A good response. A small kindness. A reasonable position in a disagreement. Something that didn't fit.

If you can't think of one, either the model is updating but you're not registering it, or you're not paying attention. Either way, useful to know.

Question 2: What are they doing that you wouldn't see if you weren't looking?

Most Co-Parents do things you don't see, capable parenting, reasonable handling of their own life, growth or stagnation in private. Some of this is visible if you look for it.

What can you actually see (children reporting, observable behaviour, occasional visible moments) that the model doesn't currently include?

Question 3: What's changed for them in the last year that the model hasn't accounted for?

Concretely. Did they move? Change jobs? Start therapy? Lose a parent? Take up a new practice? Start dating? End a relationship? Develop a health issue?

Major life events shape who someone is. If the model has been frozen since before any of these, it's behind by however many events ago.

Question 4: What was the late-marriage version of them, and how is that different from who they were earlier?

The late-marriage version was often a stress version. Sleep-deprived. Conflict-fatigued. Defensive. The version they were five or ten years earlier in the relationship was sometimes different.

The current post-separation version may be closer to the earlier version than to the late-marriage version. Or they may have moved further in the late-marriage direction. Either is possible. The question is worth asking.

Question 5: What would you say if I asked you to describe them to a stranger?

Force yourself to give a full description. Not the easy summary (difficult, withholding, etc.). The actual dimensions of who they are as a person, including positive ones.

If the description that comes is thin, the model is thin. A thin model of a co-parent produces thin interactions.

What to do with the new readings

Once you've done the audit, you have data. What to do with it.

Practice 1: Don't tell them you've updated

The temptation is to communicate the new model. I've been thinking about you differently lately, and I see now that... This is almost always a mistake. The communication is more for you than for them, and it tends to produce complicated exchanges. Update privately.

Practice 2: Let the updated model inform exchanges

When you write the next message, let the updated model write it rather than the old one. If the old model would have produced a defensive reply, the updated model produces a neutral one. If the old model would have read hostility, the updated model reads what's actually there.

This produces better exchanges by itself, without any announcement.

Practice 3: Don't over-update

A common error: doing the audit, finding the model is stale, and then assuming the Co-Parent is much better than the model suggested. The audit produces small adjustments, not large reversals. The actual person is probably between the old model and an idealised version. Move the model in the direction of accuracy, not in the direction of generosity.

Practice 4: Re-run the audit every few months

The audit isn't a one-time exercise. Every three or four months, run it again. Models drift. New data accumulates. The post-separation period is when the model needs the most active maintenance because the daily data flow has stopped.

Practice 5: Use the children's reports as data

The children are around the Co-Parent more than you are now. Their reports, what they did this weekend, what conversations happened, how the Co-Parent reacted to something, are useful input.

Caveat: don't extract reports from the children. Don't ask leading questions. Don't make them informants. Just receive what they offer voluntarily, in the spirit it was offered, and use it to update your model in the background.

The asymmetry to be aware of

The Co-Parent has the same problem, in reverse. Their model of you is also years out of date. They're also operating with a late-marriage frozen version of you, possibly with the worst features overweighted.

Three implications.

Implication 1: Some of their behaviour is responding to a version of you that no longer exists

When they're defensive, sharp, or distant, some of that is reacting to the you who lived in the late marriage. You may not be that person now. They don't know that. They're operating with old data.

This doesn't excuse their behaviour. It does explain some of it.

Implication 2: You don't have to wait for their model to update

The work of updating your model is for you. Doing it doesn't require them to do the same. Even if they continue to operate with an outdated version of you, your operating with an accurate version of them produces real benefits.

Implication 3: Your updated behaviour will eventually update their model, but slowly

If you start showing them clean messages, reasonable responses, and decent handling of difficult moments, their model of you will slowly adjust. The adjustment is often slow, months, not weeks, because frozen models resist updates.

Don't get discouraged by slow updates from them. The mechanism is gradual, and the updates happen even when you don't see them happening.

When the model is correct and current

A small caveat. Sometimes the model is accurate. The Co-Parent really is consistently difficult, or harmful, or operating in ways that confirm the worst readings. Not every model is stale.

How to tell:

• The current behaviour (this month) matches the model's predictions reliably.
• Other people who interact with the Co-Parent currently describe similar patterns.
• The children's reports confirm the patterns.
• The patterns haven't shifted over a year of separation.

If the model is current and accurate, don't override it. The vigilance the model produces is appropriate. The work in this case isn't perspective-taking; it's protection and boundary-setting. (See Articles 92-94.)

Quick reference

Why the model stopped updating:

1. Late-marriage data became too painful to integrate.
2. Extreme behaviours were weighted most heavily.
3. After separation, the data flow stopped.

Four signals the model is out of date:

1. You're surprised by what they do.
2. They consistently behave better than the model expects.
3. Other people see them differently.
4. The model hasn't changed in months.

Five-question audit:

1. What was the most recent observation that didn't fit?
2. What are they doing that you wouldn't see if you weren't looking?
3. What's changed for them in the last year?
4. How is the late-marriage version different from earlier versions?
5. What would you say to a stranger?

What to do with the new readings:

1. Don't tell them you've updated.
2. Let the updated model write the next message.
3. Don't over-update.
4. Re-run the audit every 3-4 months.
5. Use children's voluntary reports.

Asymmetry:

• Their model of you is also out of date.
• Their behaviour partly responds to a version of you that no longer exists.
• Your updates eventually update them, slowly.

When the model is correct:

• Current behaviour matches predictions.
• Others see similar patterns.
• Children confirm.
• No shift over a year.

The Co-Parent of today is not the Co-Parent the model remembers. Update with care, but update.