Turning 'Trust Me' Into a Track Record You Can Show

What each one is actually for

The Trust Equation

Four numbers, each of them a claim about what is going on inside another person's head, and each of them capable of being wrong in a direction you could eventually find out about — that is what this model asks you to produce in place of a general feeling about how a relationship is going. The four are credibility, whether what you say checks out and whether your track record supports the claims you make with it; reliability, whether you do the thing you said by the day you said it; intimacy, whether the other person feels safe bringing you something they would normally hold back; and self-orientation, how much of your attention appears to them to be on your own interest rather than theirs. The first three are summed and the fourth divides them, which is a substantive claim rather than notation: the fourth term is not weighed against the other three, it is applied to their total, so a knowledgeable, punctual, warm person can still come out low. Two things are worth knowing before you put numbers on yourself. The estimate has to be made from the other person's likely point of view rather than your own, which makes every score a forecast about somebody else and not a self-assessment — and forecasts about other people's readings of you are exactly the kind that feel obvious and turn out skewed. And this is a practitioner's framework, developed by the consultants David Maister, Charles Green and Robert Galford, with no scoring manual and no norms behind it, so a number here is not a measurement. It is a hypothesis about which of the four to go and work on, and the only thing that tests it is whether anything visibly changes in that one relationship over the month that follows.

Superforecasting

Calibration training means practicing the specific skill of attaching an honest, well-reasoned number to your own uncertainty, rather than defaulting to vague words like probably or likely that mean wildly different things to different people. You take a real question with an eventual, checkable answer — will this project ship on time, will this candidate get the job, will this market move a certain way — and instead of a confident yes or no, you commit to a specific percentage, say 65 percent, based on what you actually know. Then, crucially, you track it: you write the estimate down with a date, and later, once the answer is known, you check how your 65-percent calls performed as a group — if events you called 65 percent likely actually happened close to 65 percent of the time, you're well calibrated, and if your 90-percent calls only came true 60 percent of the time, you've just found a specific, correctable pattern of overconfidence in your own judgment.

What you get out of pairing them

What the Trust Equation buys you is a diagnostic instead of a vague feeling — instead of people don't quite trust me as an unworkable problem, you get a specific answer about which of four levers is actually low, which turns an amorphous personal quality into something you can practice. What calibration training buys you, more narrowly than a lot of pairings on this site, is one very specific upgrade to just one of those four levers: credibility. Most people's credibility rests on how they sound — confident, articulate, certain — rather than on anything a colleague could actually go check. A tracked calibration record is something a colleague genuinely can check: not trust me, I'm good at this, but here are forty predictions I made with stated confidence levels, and here's how they actually turned out. That is a different, harder-to-fake kind of credibility than sounding sure of yourself, and it's the specific thing calibration training is built to produce.

What The Trust Equation 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.
• Self orientation inversely predicts trust practitioner report
  The degree to which a person appears focused on their own interests rather than the other party's interests inversely predicts how trustworthy they are perceived to be.

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

The Trust Equation treats credibility as something that should rest on a checkable record rather than on how convincingly someone talks — you don't just believe a person is competent because they say so with confidence; you look for evidence their claims have actually held up over time. Calibration training produces exactly that kind of record, deliberately: a dated log of specific numeric predictions next to what actually happened, which is a harder, more honest form of the same evidence the Trust Equation says credibility should be built on. Where they meet is narrow but real: one is a model naming what should count as credibility, and the other is a concrete method for actually generating it.

The principle underneath both

A record other people can independently check outperforms a record you merely assert, whether what's being checked is whether you follow through on your word or whether your stated confidence levels actually hold up against reality. Confidence that sounds convincing and confidence that is demonstrably accurate are different things, and only the second kind survives someone actually going back and checking.

What that makes the real goal

Both are aimed at replacing an unverifiable impression with something a specific person could, in principle, go check for themselves — a colleague deciding whether to trust your judgment, or you deciding whether to trust your own. The Trust Equation names credibility as the lever this applies to; calibration training is one concrete way to build the actual evidence that lever depends on, though it's worth being honest that it does nothing directly for the equation's other three terms.

Other practices that serve the same goal

Structured 360-degree feedback, where multiple colleagues rate a person's specific named behaviors rather than giving a general impression, pursues the same checkable-record objective for the Trust Equation's other levers — reliability and intimacy specifically — and has mostly observational support from organizational psychology rather than randomized trials. Prediction markets, where many people stake something on a specific outcome and the market price aggregates their collective calibration, pursue the identical judgment-accuracy objective calibration training does, at a group rather than an individual level, and have observational evidence showing they often outperform individual experts. Keeping a simple decision journal — writing down a significant decision, your reasoning, and your confidence at the time, then reviewing it once the outcome is known — is a lighter version of calibration training aimed at the same personal-credibility objective, with mostly anecdotal support from practitioners in fields like investing.

A tradition that already does this

The early Royal Society adopted the motto nullius in verba, roughly take nobody's word for it, and organized its scientific credibility explicitly around claims other members could independently check through repeated experiment rather than around the personal authority or eloquence of whoever was speaking. It's an institutional precedent, not a personal practice, for the same underlying idea both the Trust Equation and calibration training rest on: a checkable record is a sounder basis for credibility than a confident voice.

Using Calibration Training to Build the Credibility Term

If the Trust Equation tells you credibility is low — people hedge on your recommendations, ask for a second opinion, or seem to trust your confidence more than your content — the fix isn't to sound more certain, which usually reads as exactly the self-orientation the model warns against. Start a simple calibration log for the specific kind of judgment your role depends on: every time you make a real prediction or recommendation with a stated confidence level, write the date, the call, and your percentage in one place. After a few months, review it honestly — not to punish yourself for misses, but to see whether your 80-percent calls are actually landing around 80 percent of the time. When it comes time to make the case for your judgment to someone else, you now have something concretely different to offer than confident tone: an actual, dated record, including the times you were wrong, which reads as more credible precisely because it isn't curated to look perfect.

Using the Trust Equation's Self-Orientation Term to Improve Calibration

One of the more common ways calibration training quietly fails is that people keep giving falsely confident numbers in front of others even after their private tracking shows they're overconfident, because sounding certain feels better in the moment than sounding hedged — this is self-orientation working directly against accurate calibration. Before you state a probability to someone else, ask the Trust Equation's question about yourself in that moment: am I picking this number because it's my honest best estimate, or because a rounder, more confident-sounding number will make me look more sure of myself? Naming your actual uncertainty out loud, I'd put this at 60, not 90, and here's specifically what I'm unsure about, lowers your self-orientation in the equation's terms — you're serving the other person's need for an accurate picture over your own need to look decisive — and it's also literally the more accurate, better-calibrated answer, since owning real uncertainty is what calibration is asking you to do in the first place.

A Client Asks If the Launch Will Succeed

A client asks you directly whether their product launch next month will hit its target, and every instinct says to sound sure — a confident yes, this is going to work is what they want to hear, and it's what would make you look like the expert they're paying for. You've been keeping a calibration log for two years, though, and you know your instinct toward confident-sounding answers has historically run ahead of your actual accuracy. You say instead: I'd put this around 65 percent, mainly because of how the pre-orders are tracking against last year's comparable launch. I keep a record of calls like this — over the last two years my 60-to-70-percent range has landed right around 65 percent actual, so I trust this estimate about as much as I trust any of them. The client doesn't get the reassuring certainty they might have wanted, but they get something a confident guess never gives them: a specific number, a stated reason, and a track record they could, if they wanted to, actually go check.

When this is worth reaching for

Your professional relationships depend on your judgment being trusted — consulting, analysis, medicine, managing people who regularly get forecasts or recommendations from you and have to decide how much weight to give them. It has much less to offer the trust-building most personal relationships need, where the thing at stake usually isn't the accuracy of your predictions.

Where this does not apply

In close personal relationships, leaning on explicit probability language where warmth or reassurance is what's actually needed can read as cold or evasive rather than trustworthy — the intimacy term in the Trust Equation is built through emotional presence, not through hedged percentages, and calibration talk does nothing for it and can actively work against it. It's also the wrong tool in situations that genuinely require a clear, unhedged answer — a safety-critical decision, an urgent yes-or-no call — where visible hedging can read as low reliability rather than honest calibration, even when the hedge is the more accurate answer.

What the evidence does and does not support

This is a narrower intersection than most pairings on this site, and it's worth being direct about that rather than dressing it up. The Trust Equation itself is a practitioner framework built from consulting experience, not a research-tested model with trial data behind it; its four terms are a useful way of thinking, not a measured finding. Calibration training, by contrast, has real research behind it — Philip Tetlock's Good Judgment Project tracked thousands of forecasters over years and found that deliberate calibration practice measurably improves the accuracy of probability judgments. What connects them here is a single, specific claim — that a tracked calibration record can serve as evidence for the credibility term in an unrelated trust model — and that specific claim has not itself been tested by anyone. Use this pairing as a plausible application of a well-supported skill to a widely used but informal model, not as a proven combination.