00 · The opening argument
We keep asking the wrong question.
Ask “is she trustworthy?” and it sounds like a question about a person. Usually it is a question about a bet. Should I lend her the money? Tell her what happened? Put her name on the project? Let her pick up my child? The person is the same in each case; the bet is not.
Most of us learned trust as a verdict that arrives as a feeling. Someone feels warm, familiar, impressive, safe. But warmth mostly tells you how easy someone is to be around. Affection tells you that you care. Status tells you that others defer to them. Access, being inside the family, the team, the group chat, tells you they were let in. None of these tells you whether they will do what they said on a Thursday when keeping their word costs them something.
Two words in that definition do most of the work. Vulnerability: if nothing of yours is at stake, you do not need trust at all, only politeness. Expectations: these are forecasts about future behaviour, and forecasts can be checked against what happens.
So trust is a verb in two senses. It is something you do: lend, delegate, confide, wait, rely. And trustworthiness is something others show by doing, over time, in conduct you can observe. Your trust is the size of the bet. Their behaviour is the evidence. Most trouble begins when the bet grows faster than the evidence, or when we stop looking at the evidence because the feeling is so loud.
The hard part is that we can see behaviour but not minds. The same unanswered message might mean indifference, a sick parent, a dead phone or a deliberate signal. This guide offers one discipline above all others: keep what you observed separate from the story you added, then choose a response that does not require you to be certain about the story.
You cannot audit a heart. You can notice whether the promise was kept.
I · Domain
Trust with what?
Meera trusts her brother with her flat keys and never with her tax return. She trusts her accountant with the tax return and never with her heartbreak. Fiction None of this is a contradiction. It is what calibrated trust looks like.
Benevolence and integrity live inside other people's heads, so for daily use this guide translates them into things you can watch. Interpretation
- Reliability. Commitments made against commitments kept, and, just as telling, what happens when one cannot be kept. Do they warn you early or let you find out?
- Candour. Do they tell you things you will not enjoy hearing, including about their own mistakes?
- Competence. Can they actually do this thing? A devoted friend may be a hopeless accountant. That is a fact about skill, not loyalty.
- Regard. When your interests and theirs diverge and nobody is watching, what happens? This is the hardest to observe and the easiest to over-read, so it deserves the most humility.
The practical question is never “do I trust them?” but “what am I about to hand them, and what have I seen them do with something similar?”
II · Observation and inference
The camera test.
Ask of any sentence about another person: could a camera have recorded this? “He replied after nine hours” passes. “He's losing interest” does not; it is a story stitched onto the footage. Stories are not forbidden. We cannot live without them. The point is to know when you are telling one.
Your feelings belong on the bottom step too. “I felt humiliated when the mistake was announced in the group chat” is a true observation, about you. What it does not establish is that humiliation was the aim. Feeling hurt is evidence that something matters; it is not, by itself, evidence of intent. Interpretation
Two-minute calibration: camera or story?
Decide for each line, then compare. Some are honestly mixed. Nothing is recorded or scored.
-
“She said she'd call on Friday. She called on Monday.”
Reveal
Camera. Two recordable events and a gap between them. What the gap means is a separate question.
-
“She doesn't value my time.”
Reveal
Story. A claim about her values, inferred from conduct. It may be true; it cannot be filmed.
-
“He presented the slides I made and didn't mention me.”
Reveal
Mostly camera, with a catch. “Didn't mention me” is only an observation if you heard the whole presentation. Second-hand accounts are not footage.
-
“He always takes credit for other people's work.”
Reveal
Story. “Always” is a generalisation and “takes” implies intent. Ask: how many times, seen by whom?
-
“My manager rolled her eyes when I spoke.”
Reveal
Mixed. Eyes moving is recordable. “Rolled her eyes” already contains a reading of contempt. She might also have been glancing at a notification.
-
“They apologised, and did the same thing two weeks later.”
Reveal
Camera. Two events in sequence. This is exactly the kind of observation that should carry weight.
-
“The apology wasn't sincere.”
Reveal
Story. Sincerity is inside them. What you can observe is what the apology named, what it offered and what happened afterwards.
-
“I felt anxious when I didn't hear back.”
Reveal
Camera, of you. Your own feeling is a legitimate observation about your experience. It tells you something matters, not why they were quiet.
III · Time
Patterns outweigh one-offs. Mostly.
A single lapse is weak evidence because so many ordinary causes produce one: a migraine, a bus, a forgotten calendar alert. A pattern across weeks and situations narrows the possible explanations. That is why the people who know someone longest are often slowest to condemn and quickest to notice a genuine change. Interpretation
- kept
- missed
- warned in advance
Four questions turn a hunch into a pattern worth acting on: How many times? Across how many situations? Did they notice or own it? Did anything change after it was raised?
Where the rule breaks
- Severity. One act that puts your safety, body, money or livelihood at serious risk can be reason enough for distance. You do not need a sample size to leave a burning building.
- Concealment. A lie told to cover a lapse is often more informative than the lapse itself, because it is a second, deliberate act.
- Where you stand. People can be impeccable towards those with power over them and careless towards those without. A pattern is only your evidence if it was observed from your position.
- Change. People do change. An old pattern followed by months of new, consistent conduct is a new pattern.
- Manufactured consistency. Someone who intends to exploit can be scrupulous in small things to earn a large bet. A useful guard: stakes should grow at the speed of evidence. Notice when what you are being asked to hand over grows faster than what you have seen. Interpretation
IV · After the breach
Repair is not the same as accountability.
When trust breaks, four different things get blurred into one word, “sorry”. Separating them makes both forgiving and refusing easier to do well. Interpretation
- Accountability
- The consequences that follow from what happened: fixing the damage, changing the procedure, perhaps losing a role. Owed whether or not the relationship continues.
- Repair
- The joint work of rebuilding the relationship. It needs both people's willingness, so it cannot be demanded by the person who broke it.
- Forgiveness
- Your own release of resentment. It can happen without reconciliation, and it can take as long as it takes.
- Reinstatement
- Handing back the same access as before. Often confused with forgiveness. You can forgive someone and still not return the keys.
Taken together, these suggest why “I'm sorry I made a mistake” lands more easily than “I'm sorry I lied”: a skill can be fixed with practice, while an integrity breach invites the question of what else was hidden. For everyday purposes, words are the cheap part of repair. Watch for the costly parts: Interpretation
- It names the specific act, not “if anything I did upset you”.
- It does not ask you to manage their guilt or reassure them they're still good.
- It offers something concrete: fixing, compensating, correcting the record in the same room the damage happened.
- It changes the condition that made the breach possible.
- It accepts your timeline, including “not yet” and “no”.
- The behaviour holds when they think nobody is checking.
A caution against the opposite error: demanding a flawless apology can become its own power move, and many families and cultures apologise mostly through acts rather than words. A soup delivered without comment can be more repair than a well-structured paragraph.
V · Distance
Boundaries are not punishments.
A boundary describes what you will do to protect something you care about: “If plans change after six, I won't be able to come.” A punishment aims to make the other person feel a cost: “Let's see how she likes being cancelled on.” Both can involve the same action. What separates them is purpose, proportion and whether a way back remains. Interpretation
Three quick tests
- Would I still do this if they never found out I'd done it?
- Is it sized to the risk, or to my anger?
- Does it leave a route back if the evidence changes?
Reasons people may reasonably choose distance
No verdict about someone's character is required to step back. Some sound reasons:
- The unreliability costs them little and costs you a lot.
- The relationship needs constant vigilance you cannot sustain.
- One serious breach made the risk too large for this season of your life.
- Your values or directions have diverged, with no one at fault.
- You simply no longer want to, and owe no one a trial.
Distance can also be partial. You can stop lending someone money and keep having tea with them. And the same freedom runs the other way: others may step back from you without granting you a hearing. Respecting that is part of what makes your own boundaries credible.
VI · Hierarchy
Trust under unequal power.
When an intern trusts a manager's “my door is always open”, the intern carries the risk. If the trust is misplaced, the manager loses a little goodwill; the intern may lose a reference, a contract or a visa sponsorship. Trust is never placed on level ground. Interpretation
Two practical consequences follow. Interpretation
The burden of demonstration should sit with the more powerful. “Trust me” from the person who controls your appraisal is a request, not evidence. What counts as evidence is observable: decisions explained, credit named publicly and specifically, promises given with dates and kept, bad news met with questions rather than blame, and mistakes corrected in the same room they were made in.
Silence is not consent, and compliance is not trust. People with less power often look trusting because disagreement is expensive for them. A leader who reads a quiet meeting as a vote of confidence may be reading fear. The test is what happens to the first person who raises an inconvenient fact.
The fair counterweight: people with power have real constraints, including confidentiality, budgets and decisions made above them. Not every unexplained decision is a betrayal. The question is whether they explain what they can, and say plainly when they can't.
VII · Consent and the limits of evidence
Trust is not access.
Being trustworthy does not entitle you to anyone's trust. Trusting someone does not entitle you to their time, their stories or their body. Trust is given; it cannot be extracted. Interpretation
Watch for requests to prove trust by handing over access: your phone, your location, your passwords, your silence about something that worries you. Notice the direction of travel. The request narrows your choices while being framed as closeness. You may say no and still care about the person asking.
And accept the hardest limit: you will never have certainty about another mind. Trust is precisely the decision made under that uncertainty. Over-checking corrodes relationships; under-checking lets them corrode you. The aim is not certainty but calibration, which includes checking your own record. When you have read people wrongly before, in which direction did you err?