---
author: "[[Bob Marshall]]"
tags:
  - blogpost
publish: 2026-08-23
sources:
  - https://flowchainsensei.wordpress.com/2026/08/23/the-flakes-you-never-notice/
---
## The Flakes You Never Notice

At a party a few years ago, someone told me they were going to introduce me to a friend of theirs who ran a studio. Not vaguely – specifically. They said her name. They said she’d love what I was working on. They took out their phone in front of me, which is the universal gesture for _this is now underway_.

Nothing happened. No introduction, no email, no follow-up. And here’s the part worth dwelling on: I didn’t notice. Not for years. I ran into them again, was pleased to see them, thought of them as one of the good ones. It only surfaced later, when I was thinking about someone else entirely, and the memory arrived with a small click.

That person had never done anything they said they’d do. Not once, in a decade of acquaintance. And if you’d asked me to name the flakes in my life, they wouldn’t have made the list.

### Nothing Happening Makes No Sound

We’re built to notice events. A friend who cancels an hour before dinner produces an event: a text, a scramble, a wasted booking, a flare of irritation you can point at. A friend who never gets as far as making the dinner plan produces nothing at all. There’s no moment. There’s just an absence where a thing might have been, and absences don’t announce themselves.

Psychologists have a name for the underlying handicap – _the feature-positive effect_. Across six experiments, Newman, Wolff and Hearst (1980) found that people learn to spot a pattern far more readily when it’s marked by something being _there_ than by something being missing, and concluded with the nicely deflating observation that organisms struggle to use non-occurrence as a cue. Undelivered promises are made of exactly the same material as things you never thought about in the first place. To tell them apart you’d need a ledger, and nobody keeps a ledger on their friends.

### Every Single Miss Has a Good Explanation

The other reason the pattern stays hidden is that it never presents as a pattern. It presents as a series of unrelated incidents, each one perfectly accounted for. They were moving house. Their child was ill. Work went insane in March. The thread got buried. They were about to, actually, last week.

Every one of these is true. That’s what makes them so effective at concealing the shape of the thing. To see the pattern you’d have to hold twelve separate reasonable explanations in your head simultaneously and recognise that twelve reasonable explanations in a row is not itself reasonable. We’re not good at that. We evaluate excuses one at a time, and one at a time, they all pass.

There’s a lovely echo of this in the planning-fallacy research. Buehler, Griffin and Ross (1994) found that people don’t just underestimate how long things will take – they actively discount their own past record while doing it, treating each previous overrun or miss as a one-off caused by circumstances that won’t recur. Flakes do this to themselves; we do it to them on their behalf.

### You Quietly Stop Asking

This is the mechanism that really buries it.

After a few unreturned favours, you adjust – not consciously, and not with any bad feeling. You just stop routing anything important through that person. You don’t ask them to pick up the keys. You don’t put them down for the slot that matters. If you need the thing by Friday you ask someone else, and you don’t experience this as a judgement, only as a small practical instinct about who to text. (Also applies to e.g. suppliers).

And now the failures stop, because you’ve stopped creating opportunities for them. Their record improves. You’ve reduced your sampling rate to zero and mistaken the resulting silence for reliability.

This is a well-documented trap, and it cuts both ways. Denrell and March (2001) called it the hot stove effect: because we stop sampling the options that disappoint us, our errors of underestimation never get corrected, while our errors of overestimation do. Denrell (2005) then applied the same logic to people, showing how the simple fact that we keep interacting with those we like – and quietly withdraw from those we don’t – is enough on its own to produce systematic distortions in what we believe about them. Withdrawal is a form of not-looking. It feels like a conclusion, but it’s actually the end of the enquiry.

Meanwhile the work has gone somewhere: usually to you, or to the two or three people everyone else has also quietly decided are the ones who actually do things.

### Warmth Gets Counted as Reliability

Chronic flakes are, in my experience, disproportionately lovely. Not as camouflage – I think the causation runs the other way. The same expansive, generous, present-tense enthusiasm that makes someone a delight to talk to at eleven at night is the thing generating promises their Thursday self will never honour.

Because the promise is sincere. That’s the part people get wrong when they call it dishonesty. At the moment of speaking, they fully intend it. They’re just making a forecast about a future person – a rested, unhurried person with a clear diary and nothing else on – and that person does not exist and has never existed. They’re not lying to you. They’re wrong about themselves, reliably, in the same direction, forever.

And when someone is warm, we file them under _good_, and _good_ is an undifferentiated bucket. Nisbett and Wilson (1977) showed how comprehensively a global impression bleeds into judgements of specific attributes – and, more unsettlingly, that people had no idea it was happening to them. Kind, generous, funny, _reliable_: it takes deliberate effort to keep those on separate axes. Most of us don’t bother.

### The Apology Closes the Loop, Without Closing the Loop

When it does surface, the apology is usually excellent. Disproportionate, even. Real remorse, no defensiveness, a proper account of what went wrong.

And that feels like resolution. Something happened, it was addressed, we’ve moved on. But nothing has actually been delivered. The emotional debt is discharged; the practical one is still sitting there. You come away from the conversation feeling better about a situation that is materially identical to how it was before, and – this is the bit – with the counter reset to zero.

The one honest defence of all this is that the apology probably is doing real work. Schweitzer, Hershey and Bradlow (2006) found that trust damaged by _unreliability_ can be fully repaired, provided the damage wasn’t compounded by deception. That’s the crucial distinction, and it’s why flakes get away with it so gracefully: they aren’t lying to you. There’s nothing to forgive in the way there would be with a lie. There’s just nothing there.

### Where the Cost Actually Lands

Here’s why any of this matters, given that no single instance costs you much.

You don’t notice the individual flakes, but you do notice their sum. It simply doesn’t arrive labelled. It arrives as a vague sense that things you set in motion tend not to arrive. That projects stall in the middle for no clear reason. That you seem to be carrying more than your share and can’t quite say why, or by whom. That you’re bad at making things happen.

The unreliability gets attributed to the world, or to you, rather than to the specific handful of people it belongs to. That’s the real damage – not the missing introduction, but the slow accumulation of a belief about how much friction there is in life, when actually there’s a normal amount of friction and a few named individuals.

### The Loudest Flake Is Rarely the Flakiest

One last thing, which I find useful. The person you’d name if asked – the dramatic canceller, the eleven-o’clock apologist, the one who bails on the day – is a mid-tier flake at worst. They got far enough into the process to have something to cancel. They made a commitment concrete enough that failing it registered on your calendar.

The serious cases never reach the calendar. They evaporate upstream, at the intention stage, where there’s nothing to fail against. The flakiest people you know have a spotless record with you, because the record only has room for things that got far enough to break.

### What to Actually Do

Not much, is the honest answer, and definitely not a confrontation. Nobody has wronged you. But two small things help.

#### Close the Gap Between the Promise and the Artefact

When someone says they’ll send the introduction, the move is to make it happen in the room: _amazing – shall we do it now, while we’re both here?_ Not as a test. Enthusiastic people are at their most capable during the enthusiasm, and it decays fast. Where that isn’t possible, the next best thing is to pin the vague intention to a specific trigger before you part – a date, a place, a moment. This is the whole substance of Gollwitzer’s (1999) work on implementation intentions: ‘I’ll get round to it’ and ‘I’ll do it on Sunday morning after breakfast’ are not variations in wording, they’re variations in outcome.

#### Keep the Axes Separate

Someone can be a wonderful friend and a bad conduit for tasks. Those are different columns, and putting them in the same one is how you end up either resentful of good people or dependent on the wrong ones._You’re allowed to love someone’s company and route nothing through them._

And then the uncomfortable bit, which I’ll leave here without elaborating: everything above applies to you too, somewhere, with someone. There’s a person out there who doesn’t know they’ve never once received anything you said you’d send – because nothing arriving makes no sound, and they’ve long since stopped asking.

### Further Reading

Buehler, R., Griffin, D., & Ross, M. (1994). Exploring the “planning fallacy”: Why people underestimate their task completion times. _Journal of Personality and Social Psychology, 67_(3), 366–381. [https://doi.org/10.1037/0022-3514.67.3.366](https://doi.org/10.1037/0022-3514.67.3.366) The foundational study on why people are so bad at forecasting their own future selves, including the finding that we discount our own track records when making the next prediction.

Denrell, J. (2005). Why most people disapprove of me: Experience sampling in impression formation. _Psychological Review, 112_(4), 951–978. [https://doi.org/10.1037/0033-295X.112.4.951](https://doi.org/10.1037/0033-295X.112.4.951) Shows how the ordinary tendency to keep interacting with people we like, and to withdraw from those we don’t, is by itself enough to generate durable errors in our impressions of them.

Denrell, J., & March, J. G. (2001). Adaptation as information restriction: The hot stove effect. _Organization Science, 12_(5), 523–538. [https://doi.org/10.1287/orsc.12.5.523.10092](https://doi.org/10.1287/orsc.12.5.523.10092) The original statement of the asymmetry behind the ‘you quietly stop asking’ problem: avoid something after a bad experience and the mistaken belief that produced the avoidance can never be corrected.

Gollwitzer, P. M. (1999). Implementation intentions: Strong effects of simple plans. _American Psychologist, 54_(7), 493–503. [https://doi.org/10.1037/0003-066X.54.7.493](https://doi.org/10.1037/0003-066X.54.7.493) On the gap between intending and doing, and the disproportionate power of tying an intention to a specific situational cue.

Newman, J., Wolff, W. T., & Hearst, E. (1980). The feature-positive effect in adult human subjects. _Journal of Experimental Psychology: Human Learning and Memory, 6_(5), 630–650. [https://doi.org/10.1037/0278-7393.6.5.630](https://doi.org/10.1037/0278-7393.6.5.630) Six experiments on how much harder it is to learn from the absence of something than from its presence – the cognitive basis for absences making no sound.

Nisbett, R. E., & Wilson, T. D. (1977). The halo effect: Evidence for unconscious alteration of judgments. _Journal of Personality and Social Psychology, 35_(4), 250–256. [https://doi.org/10.1037/0022-3514.35.4.250](https://doi.org/10.1037/0022-3514.35.4.250) The classic warm-versus-cold interviewer study, notable less for the halo effect itself than for participants’ complete unawareness that it was operating on them.

Schweitzer, M. E., Hershey, J. C., & Bradlow, E. T. (2006). Promises and lies: Restoring violated trust. _Organizational Behavior and Human Decision Processes, 101_(1), 1–19. [https://doi.org/10.1016/j.obhdp.2006.05.005](https://doi.org/10.1016/j.obhdp.2006.05.005) Finds that trust damaged by unreliable behaviour can recover fully, but trust damaged by deception does not – which is roughly why we forgive flakes so easily and liars not at all.