Asking for Help: You Are Very Wrong About How Willing They Are

The short answer: you are very wrong about how willing they are

Flynn and Lake's 2008 study in the Journal of Personality and Social Psychology (source) is the experiment to keep in your pocket. Across multiple field studies, people underestimated how likely strangers were to help them by roughly half. Subjects asked to predict how many people they would have to approach to get three to fill out a questionnaire predicted around twenty; the real number was about ten. The cost of asking is a story you are telling yourself. The data says the story is wrong.

Why we systematically get this wrong

Three biases stack to make asking feel more costly than it is. First, we overestimate the instrumental cost to the helper — the time, the effort, the inconvenience — and underestimate the social cost they pay if they say no. For them, declining feels worse than helping; for us, being declined is what we are bracing for. The asymmetry runs in our favour and we don't see it.

Second, the spotlight effect: we believe everyone is watching us more closely than they are. The exposure of asking feels enormous from the inside and is nearly invisible from the outside. Most of the people you ask will forget you asked within a day.

Third, asking for advice is one of the only behaviours that raises perceived competence — a 2015 Brooks, Gino, and Schweitzer paper in Management Science found that people who asked for advice were rated as more competent by the advisor, not less. The shame story has the social signal exactly backwards.

Six asks that the data says will work

• Name what you need, in one sentence, before anything else. "Can I borrow ten minutes of your brain on X?" beats a paragraph of preamble. The preamble is shame trying to soften the ask; the helper hears it as ambiguity and gets more reluctant, not less.

• Ask the person whose competence you're admitting. Brooks et al's finding cuts both ways: people you ask for advice rate you higher afterwards, and they enjoy giving it. You're not imposing — you're handing them a small status reward.

• Specify the cost in time. "Fifteen minutes this week" is a bounded ask; "can you help with my project" is an open-ended hostage situation. People say yes to the first and avoid the second.

• Give them an easy exit. "No worries if not" or "happy to ask someone else" reduces their social cost of declining. Paradoxically this raises the yes rate — they know they're choosing to help, not being trapped into it.

• Ask in writing for non-urgent help. A message lets them answer on their schedule, which lowers the cost they feel. In-person or call is for things that genuinely need a synchronous answer.

• Close the loop after. "I used your suggestion and here's what happened" is the single highest-return social move available. It costs you a sentence and makes the next ask of the same person dramatically easier.

Why this pays double if you have ADHD or RSD

ADHD brains genuinely need to ask more often — executive dysfunction means more questions, more sanity checks, more "is this on track" pulses. And rejection-sensitive dysphoria turns each ask into something that feels physically expensive. The two pressures meet in a bad place: you need help more, and asking costs more. The shame-driven response is to ask less, which is exactly the wrong direction.

What the Flynn data does, if you let it in, is reset the prior. The next ask will probably get a yes. The yes you are bracing against isn't the median outcome — it's the worst-case outcome you've trained yourself to expect. Asking five times this month with a yes rate that's quietly closer to 70% than 30% is the cheapest way to retrain the prior. The brain updates on data; it doesn't update on pep talks.

Where it fails (and how to repair)

• Asking the wrong person to protect yourself from the right one. Sometimes the lowest-shame person to ask is also the least able to help. The cost of asking the right person is high once; the cost of asking the wrong one is dragged out over weeks. Pay the one-time cost.

• Asking too late. Shame compounds with delay — week-two-of-being-stuck is much harder to confess than day-two. Set a private rule: if you can't move forward in 24 hours, the next move is to ask.

• Burying the ask. Long messages, four caveats, then the question. The helper has to dig. Move the ask to the first sentence; put the context underneath.

FAQ

Isn't this just imposing on people?

The Flynn data and the Brooks data say the opposite. People decline less than you expect when they decline at all, and the ones who help rate you higher afterwards. The model in your head where every ask is a debt is mostly an artefact of shame, not a reading of how the social transaction actually goes.

What if they say no?

Then you know — which is information you didn't have. The other thing to know is that "no" rarely means "and I think less of you for asking". It usually means they're busy, or it's not their thing, or they think someone else is better placed. Treat a no like an out-of-stock notice, not a verdict.

Doesn't asking too much make me look incompetent?

Asking too much from the same person, on the same topic, with no progress between asks, can. The 2015 Brooks paper looked at single asks for advice and found them competence-raising — but it doesn't extend infinitely. The fix is to vary who you ask, close the loop on what you got, and demonstrate progress between asks.

How do I ask my boss for help without it being a flag?

Frame it as a decision check, not a rescue. "I'm taking approach X to Y; can I get five minutes to make sure I'm not missing something?" reads as judgment and initiative. "I don't know what to do" reads as a flare. Same situation, different signal.

What about asking strangers — online forums, cold emails?

Flynn's study used strangers and the help rate was still about double what subjects predicted. Cold asks work better than the asker thinks, especially when the ask is specific, time-bounded, and the helper can see exactly what would count as success. Generic asks get generic ignores; specific ones get answers.