Using Collaborative Problem Solving to enhance Superforecasting

What each one is actually for

Collaborative Problem Solving

The first move in Collaborative Problem Solving is an information-gathering move, and it exists because the adult's private theory about why the behavior keeps happening is usually wrong. Ross Greene, the psychologist who developed the method, found that adults arrive at these conversations already holding an explanation — he's testing me, she's tired, they want attention — and that the explanation is a guess wearing the clothes of a fact, formed entirely from the outside and never once checked against the person who actually has the problem. So the method makes you check it. In a calm moment, well away from the incident, you name one specific unsolved problem in neutral, concrete terms — "getting started on homework right after school has been hard" — and then you ask what is up with that, and you stop talking and listen, including when the answer is nothing like the one you were carrying in. Only after the real obstacle is on the table do you state your own concern plainly, and only then do the two of you build a solution that addresses both. What makes the method work is not the solution step but the willingness to be wrong at the start: an adult who walks in certain they already know why gets a conversation that agrees with them and a plan that quietly fails a week later, and an adult who walks in genuinely uncertain gets the one piece of information the whole thing depended on and could not have supplied for themselves.

Superforecasting

Calibration training, drawn from Philip Tetlock's research on what separates accurate forecasters from inaccurate ones, teaches you to make explicit, numeric probability estimates about real-world questions and then track how often those estimates turn out to be right. If you say you are seventy percent confident about something, calibration means that across many such seventy-percent claims, you are actually right about seventy percent of the time, not ninety, not forty. Getting there requires two specific habits: breaking a large uncertain question into smaller, more checkable sub-questions rather than pronouncing on the whole thing at once, and keeping score, writing your estimates down before you know the outcome and reviewing them afterward against what actually happened. Picture it directly: instead of just feeling sure a project will finish on time, you write '70% confident we finish by Friday' in a note before the week starts, and when Friday arrives you actually go back and check whether that felt-sure seventy percent was closer to the truth than your gut alone would have been.

What you get out of pairing them

What Collaborative Problem Solving buys a parent is the end of a specific trap, fighting the same battle over and over with a child while quietly assuming you already understand why it keeps happening, when the actual reason has never once been asked about directly. What calibration training buys anyone making real decisions under uncertainty is the same relief applied to their own mind: the end of trusting a felt sense of certainty that has never once been checked against how often that feeling has actually been right. Neither practice is really about the child or the forecast. Both are about closing the same gap, between what you assume you already know and what you have actually gone and found out, whether the unknown thing is sitting across the table from you or living inside your own head as an unexamined hunch.

What Collaborative Problem Solving rests on

• Externalizing makes pattern visible randomised trials
  Putting an internal pattern into an external, structured form makes it visible and available for examination instead of remaining implicit and merely enacted.
• Parts have positive intent clinical practice
  A behavior that looks like misbehavior is usually a mistaken strategy aimed at meeting a legitimate underlying need rather than an expression of badness itself.
• Collaborative solution generation increases buy in observational studies
  Jointly generating a solution with the person who has the problem produces more durable follow-through than an imposed solution.

What Superforecasting rests on

• Decomposition improves estimate accuracy observational studies
  Breaking a complex uncertain question into smaller more tractable sub-questions produces more accurate estimates than judging the whole question at once.
• Explicit probability estimates improve calibration randomised trials
  Forcing a numeric probability rather than a vague verbal hedge makes a prediction falsifiable and trainable.
• Outcome feedback corrects judgment bias randomised trials
  Regularly scoring predictions against actual outcomes and reviewing the misses corrects systematic biases in judgment over time.

Where they actually meet

It matters to say plainly that the two techniques do not share a surface, nobody is asking a seven-year-old for a probability estimate, and nobody is running a three-step empathy conversation with a spreadsheet. What they share sits one level down. Plan B's opening question, 'what's up', is an act of externalizing: a child's vague, unexamined internal state gets turned into specific, sayable words instead of staying an implicit thing an adult simply guesses at. A calibration estimate is the identical act aimed at your own mind: a vague, unexamined felt sense of confidence gets turned into a specific, checkable number instead of staying an implicit hunch you simply trust. Different material, a child's motive in one case, your own certainty in the other, but the same underlying refusal: stop treating an assumption as good enough.

The principle underneath both

A claim, about why someone is behaving a certain way, or about how likely something is to happen, only becomes something you can actually work with once it has been made explicit and external rather than left as an internal impression. An impression cannot be checked, corrected, or discussed; a stated concern or a stated number can be. Both practices exist because the unexamined internal version, of a child's reason or your own confidence, is where errors hide, and putting it into words or numbers is what drags the error into view.

What that makes the real goal

Making the implicit explicit is what a Plan B conversation and a calibration estimate are both doing, and both are aimed at closing the same specific gap: between what you assume to be true and what is actually true, before you act on the assumption. That target is not owned by parenting or by forecasting. It applies anywhere a person is confidently acting on a read of a situation, another person's motive, their own certainty, that has never actually been tested against reality.

Other practices that serve the same goal

Motivational interviewing, a counseling approach with substantial randomized-trial support for behavior change, pursues the identical objective in conversations with resistant adults: instead of assuming someone who has not changed a behavior simply lacks motivation, the practitioner reflects back what they are hearing and asks open questions to surface the actual, specific ambivalence underneath the resistance, the same 'what's up' move Plan B uses on a child, aimed at an adult instead. On the judgment side, the decision researcher Gary Klein's premortem technique pursues the same objective before a decision is even made: instead of assuming a plan will work, the team explicitly imagines it has already failed and writes down, specifically, why, forcing hidden assumptions into visible, checkable form before the outcome is real, which is calibration's same discipline applied earlier, to a plan rather than a forecast. Neither tradition was built with the other in mind, and neither was built with Collaborative Problem Solving or Tetlock's research in mind either; all four independently landed on the same lever, make the assumption explicit before you act on it, aimed at whatever that field actually needed examined.

A tradition that already does this

Functional behavior assessment, a standard tool in applied behavior analysis, already treats a child's problem behavior the way a forecaster treats an uncertain claim: as an explicit hypothesis to be tested against actual data rather than assumed. A practitioner proposes a specific reason the behavior is occurring, gathers observations to check it, and revises the hypothesis if the data does not support it, the same gather-information-then-test-and-revise loop calibration training runs on world events, run instead on a single child's behavior. It predates both Greene's model and Tetlock's research and shows the same combination, treat an assumed read as a testable claim, was already useful enough to become standard practice in a third field.

Using Collaborative Problem Solving to enhance calibration training

Most people skip straight from a question to a number, guessing seventy percent on instinct without ever asking what is actually driving the uncertainty, which is exactly the mistake Plan B is built to prevent on the parenting side. Before assigning a probability to something genuinely uncertain, a project finishing on time, a decision going your way, borrow Plan B's information-gathering step and ask your own version of 'what's up': write down two or three specific things you actually do not know yet that are driving the uncertainty, the same way you would ask a child what specifically makes the shoes hard rather than assuming you already know. Only after naming those specific unknowns do you assign the number. The forecast that comes out the other side is still a guess, but it is a guess built on named unknowns instead of an unexamined hunch, which is the entire difference Plan B is trying to make between an assumed reason and a found one.

Using calibration training to enhance Collaborative Problem Solving

Plan B ends with a jointly generated solution, but Greene's model does not build in a habit of actually checking, later and specifically, whether that solution held up, which leaves a parent running the same conversation repeatedly without ever noticing whether their own sense of which solutions tend to work is accurate. Borrow calibration training's discipline for the review phase, never for the conversation itself. After a Plan B session produces a solution, privately write down a rough number, how confident are you, honestly, that this specific solution actually holds up for the next two weeks, before you find out. Then actually check back at the two-week mark and score yourself. Over several rounds of this, a parent typically discovers a specific pattern in their own miscalibration, often overconfidence in solutions that sound clean in conversation but assume more follow-through than a child that age can actually sustain, which is information no single Plan B conversation would ever surface on its own.

Two weeks after the shoes conversation

The Saturday conversation about the shoes produced a solution: switch to slip-ons, no laces, problem solved, and it felt clean enough that you privately wrote '85% confident this fixes the morning meltdowns' in a note on your phone before you let yourself forget the number. Two weeks later you actually go back and check. The slip-ons helped, mornings are calmer, but not eighty-five-percent calmer; there is still a meltdown most Thursdays, and when you look closer you realize Thursday is early gym day and the actual driver was never the laces alone, it was rushing. You would not have caught that specific gap between your confidence and the outcome if you had not written the number down beforehand and actually gone back to check it, the same way you would never have known the laces mattered at all if you had not asked instead of assumed.

When this is worth reaching for

Plan B conversations already happen regularly here — the method itself is fluent. What's off is quieter: a private hunch about which solutions will actually stick, and that hunch turning out wrong more often than it should, in a pattern nobody has named yet. This work belongs entirely in the private planning and the later review of a solution, never inside the conversation itself.

Where this does not apply

Never bring calibration language, or anything that resembles scorekeeping, into an actual Plan B conversation or anywhere near a child who is currently dysregulated; Greene's own model is explicit that Plan B only works in a calm proactive window, and introducing a probability estimate into a live emotional moment would rupture the empathy step rather than support it. And if you find yourself assigning a child's motives a confidence score in your head during the actual conversation instead of genuinely listening, you have stopped doing Plan B at all; the calibration habit belongs strictly before and after the conversation, never inside it.

What the evidence does and does not support

Both halves are independently well-supported and this pairing does not borrow strength from either to prop up the other. Collaborative Problem Solving has randomized-trial evidence showing reduced explosive behavior and reduced use of restraint in clinical settings, and one head-to-head trial found it comparable to established parent training approaches; most home-setting evidence remains clinical and observational rather than trial-based. Calibration training has solid trial support from forecasting-tournament research showing that decomposition and explicit numeric estimates, scored over time, measurably improve accuracy. Neither literature has anything to say about the specific move this page proposes, using a private confidence score to review a Plan B solution's durability, which is this page's own extrapolation from Tetlock's forecasting research, built for predicting geopolitical events, into a domestic setting his studies never examined. Treat that specific application as a reasonable borrowed habit, not a finding.