People Will Help You More Than You Think

There is a message in your drafts you are not going to send. Maybe it asks someone to check in on you every Sunday about the thing you keep abandoning. Maybe it asks a friend who runs to let you tag along, badly, for a few weeks. Maybe it asks someone whose work you admire to look at yours for twenty minutes.

You have rewritten it twice to sound casual. You have added an escape hatch so generous that saying no requires no effort at all. And you are still not going to send it, because you can picture the small pause before they reply, and the polite version of no, and how it would feel to have asked.

Run the numbers on that picture, though, and it turns out to be wrong in a specific and measurable direction.

The Size of the Error

Francis Flynn and Vanessa Lake ran six studies on this, combining laboratory experiments with natural field settings where participants walked up to real strangers and made real requests. Before asking, participants predicted how many people would agree.

In the first three studies, people underestimated by as much as 50% the likelihood that others would agree to a direct request for help. Not on one unusual request. Across a range of them, in the lab and out in the world.

Half. You are standing there braced for a refusal that, statistically, is not coming nearly as often as you think.

Fourteen Thousand Strangers Later

That was 2008. The research program kept going, and Vanessa Bohns, who published the original work as Lake, later reviewed the whole thing. By then, participants in these studies had approached more than 14,000 strangers with a variety of requests.

That distinction is worth being pedantic about, because the stat gets mangled constantly. The 48% is not a success rate. It is how far off the requesters' own predictions were. Bohns describes the error as one that persists across samples and across different kinds of request, and calls it substantial.

People, in her framing, are overly pessimistic about their ability to get others to comply.

It Is a Seat, Not a Personality

The tempting explanation is that some people are just anxious about asking. Studies 4 and 5 make that hard to sustain.

Experimentally manipulating a person's perspective, putting them in the help-seeker seat rather than the potential-helper seat, was enough to produce the underestimation. Same people, different chair, different forecast.

Which means this is not a trait you have and your bolder friends do not. It is what the request looks like from where you are standing, and everybody standing there sees it the same way.

What Each Side Is Actually Looking At

The sixth study went after the reason, and the answer is a clean asymmetry of attention.

Help seekers focus on the instrumental costs of helping: the time it will take, the effort, the inconvenience you are about to hand someone. That is a real cost and you are not wrong to see it. What you discount is the other side of the ledger. Potential helpers are far more aware of the social cost of refusing, of how genuinely awkward it is to look at another person and say no.

You are running the numbers on one column. They are running them on both.

There is something slightly uncomfortable in that, and it should be said out loud: part of why people say yes is that saying no is unpleasant. This is a reason to ask well rather than a reason not to ask. Keep requests bounded, make the shape of the commitment clear, and do not use a stranger's discomfort as a lever for something big.

What This Costs You on a Goal

Most goal advice tells you to get a witness. The evidence behind it is decent: Gail Matthews' work at Dominican University, which we covered in what is an accountability partner, found that people who wrote goals down, told a friend and sent weekly progress notes finished well ahead of those who only thought about their goals.

The step nobody analyses is the one before that. Every accountability arrangement starts with somebody asking, and the underestimation effect sits directly on that step. You do not decline to have a partner. You decline to send the message, on the basis of a probability estimate that is reliably, measurably too low.

That is a strange way to lose a documented advantage: not by rejecting the advice, but by silently pricing the first move too high.

It compounds in a way worth noticing. Because you never send the message, you never collect a data point, so the estimate never updates. The imagined refusal stays undefeated on a record of zero attempts, and every month it gets a little more authoritative for having gone unchallenged. Meanwhile the person you would have asked has no idea any of this happened. From where they sit there was no request, no awkward pause, no polite no. There was a Tuesday.

The asymmetry is almost funny. You ran an entire negotiation, lost it, and were the only participant.

Worth pairing with one nuance from the other direction. Whether sharing a goal helps depends on who hears it, and Klein and colleagues found the benefit shows up with a higher-status audience, someone whose opinion you actually weigh. So the finding here is not that you should broadcast. It is that the specific person you want to ask is far more likely to say yes than the version of them in your head.

Where the Effect Stops

Three honest limits, because this is the kind of finding that gets flattened into "just ask, people always say yes."

The requests in these studies were modest and mostly one-off, made to strangers. Filling in a questionnaire, borrowing a phone, small things. Nothing establishes that people will agree to carry you for six months.

Bohns explicitly notes factors that strengthen the effect, weaken it, and reverse it. It is a reliable bias, not a law, and the review says so.

And the research was conducted in person and in the lab. What happens when you ask by message is a separate question that these studies do not answer, though it is fair to note that a text is a much easier thing to leave unanswered than a person standing in front of you. If a request matters, the evidence gives you no reason to prefer the low-friction channel.

How to Send the Thing in Your Drafts

Ask directly rather than hinting. The entire literature is about direct requests. A hint gives someone the option of not noticing, which is the option most people quietly take, and then you conclude nobody was interested.

Make it specific and bounded. "Would you be up for a five-minute check-in on Sunday nights for the next month" is a request someone can actually evaluate. "Would you help keep me accountable" is a commitment with no visible edges, and open-ended asks are the ones that get the polite deferral.

Stop pre-refusing on their behalf. The escape hatch you wrote into paragraph three is you doing their job for them. A little of that is courtesy. A lot of it is you resolving the request before they get a say.

Notice which chair you are in. When someone asks you for something small, note how you actually respond, and how little the imposition registers an hour later. That is the correction, and it is available every week for free. Nobody has ever spent a Thursday evening resenting a colleague for asking them to look at a document on Tuesday. Somebody, somewhere, is currently not sending you a message because they think you would.

If asking one specific person feels like too much to start with, the lower-stakes version is joining somewhere people are already doing this in public. That is most of what a community around goals is for, and it is what Future You is built as: your progress visible to people working on their own, so the first ask is a smaller one. It is free on iOS and Android.

Send It

You are not going to talk yourself out of the feeling. The seat produces the pessimism and knowing about the seat does not vacate it, which is why this is a finding you act on rather than one you internalise.

So take the draft, cut the third paragraph where you apologise for existing, make the request specific enough to answer with a yes or no, and send it. You will still expect the refusal. That expectation has been measured against fourteen thousand real requests and found to be too low, and there is exactly one way to collect your own data on it.