The ick comes from
arriving empty-handed

The discomfort is a signal, not a defect. When an encounter is all take and no give, you are the one who files it under "using someone" — and once it's filed there, your posture and phrasing distort, and what reaches the other side is tension and calculation. Arriving with something genuine reclassifies the encounter, and the alarm switches off. Giving also isn't pure cost: it runs on the reward circuit itself, which is why give-first is one of the few approaches anyone sustains for years.

• Direct evidence · instrumental networking feels dirty: Casciaro, Gino & Kouchaki (2014, Administrative Science Quarterly) combined experiments with field data from a law firm: networking for professional gain, versus out of genuine interest, triggered more feelings of impurity and led to less networking, and the effect was strongest for people lower in the hierarchy. Discount honestly: one measure was a cleansing-word completion task, kin to the poorly-replicating Macbeth effect (Earp et al., 2014 failed to replicate) — the finding that people report discomfort and consequently do less holds up; the "moral impurity" interpretation shouldn't be treated as settled.
• Hard evidence · giving activates the reward system: Harbaugh, Mayr & Burghart (2007, Science), an fMRI study: when money went to a charity, activity rose in the nucleus accumbens and ventral striatum (core of the dopaminergic reward system), and rose further when the giving was voluntary. Moll et al. (2006, PNAS) found the same direction independently. This is why give-first is sustainable and grinding through cold networking isn't.
• Hard evidence · reciprocity is intrinsically rewarding: Rilling et al. (2002, Neuron) scanned people playing the prisoner's dilemma: mutual cooperation activated the nucleus accumbens, caudate and orbitofrontal cortex more strongly than winning alone did.

Boundary: giving first is not ingratiation. If you give and then start a private ledger, waiting to collect, people can feel it — that's just deferred extraction. The test: if they could never repay you, would you still do this? If you can't say yes, don't dress it as generosity; asking outright is more dignified.

The mechanism is information redundancy, not headcount. You and your close friends soak in the same pool, so the marginal information they can hand you approaches zero. The line reaching another cluster — structurally, a bridge — is the entry point for anything new, not because that person is smarter but because they stand where you can't see. Yet weaker isn't uniformly better: someone too distant doesn't know you well enough to vouch. The valuable band is the middle: they remember you and know roughly what you're worth, but they've left your information pool.

• Hard evidence · large-scale causal experiment: Rajkumar et al. (2022, Science) used five years of randomised connection-recommendation experiments across roughly 20 million LinkedIn users, tracking subsequent job changes: moderately weak ties (around ten mutual connections) produced the largest increase, while the strongest ties contributed least causally. Rare genuine causal evidence in this area — hard. Boundary: it covers white-collar job search only; don't extrapolate to selling products or raising money.
• Classic, but discount it: Granovetter (1973, American Journal of Sociology) introduced the weak-tie argument; the interview data in his 1974 Getting a Job found that among people who landed a job through a contact, only about one in six came from someone they saw often, while more than half came from someone seen occasionally. A few hundred respondents, 1970s, all male professionals — read it as a direction, not a universal ratio.
• Hard evidence · strong ties do things weak ties can't: Krackhardt (1992) on "the strength of strong ties": where trust, risk, and real effort are required — vouching for you internally, helping during a crisis — strong ties clearly outperform. The two kinds of tie do different jobs.

Boundary: weak ties pay only if they genuinely cross into another cluster. Two hundred people you added in an industry group chat aren't weak ties, they're strangers in the same pool — same redundant information, minus the history. And this rule is most reliable for finding jobs and information; when you need someone to actually put in effort, go back to strong ties.

This is a repeated game, and what people evaluate isn't this particular request but your pattern. Someone who surfaces only when they need something has an unmistakable pattern, and by the third time nobody replies — not out of meanness, but because their time is being consumed in one direction. What a cheap deposit changes isn't a debt, it's your retrievability: whether your name comes to mind when they're talking to someone else. A dormant tie, meanwhile, holds two advantages at once — the trust doesn't need rebuilding, and the years apart mean your information no longer overlaps.

• Medium · dormant ties are more useful: Levin, Walter & Murnighan (2011, Organization Science) had 224 EMBA executives reconnect with people they hadn't spoken to in at least three years and ask for advice: that advice was rated more valuable and more novel than advice from their currently active contacts. Discount: self-rated, executive sample, no randomised control.
• Hard evidence · reciprocity is intrinsically rewarding: Rilling et al. (2002, Neuron): rounds where cooperation held significantly activated the nucleus accumbens and caudate. Which is why mutual-help relationships don't run on willpower, and one-directional ones can't survive — one side is collecting no reward at all.
• Counter-example (stop treating 150 as a hard number): Dunbar's number circulates everywhere, but Lindenfors et al. (2021, Biology Letters) reanalysed the original brain-size-to-group-size data and found confidence intervals far too wide to support any such precise figure. "Maintenance has a real cost, so prioritise" is fine; planning your network around a ceiling of 150 has no basis.

Boundary: the account metaphor has one dangerous use — recording every gift as a receivable and waiting to collect. Once you're keeping that ledger the flavour changes and people smell it. Safer to treat each deposit as spent: assume no return, and count any return as a surprise. Also, don't turn deposits into pestering — a weekly check-in with no substance drains the same balance.

Your forecast runs on an incomplete ledger: you can see the time a yes costs them, and you can't see what a no costs them — the social price of refusing someone to their face, which for most people outweighs the small favour. The blind spot is systematic and always points the same way, so "they probably won't say yes" is wrong as a default. That same asymmetry explains channel: email hands them a zero-cost refusal. And all three dials do one job — push the cost of yes below the cost of no.

• Hard evidence · people underestimate compliance: Flynn & Lake (2008, Journal of Personality and Social Psychology), five experiments: participants predicted they'd need to approach about 20 strangers to get 5 to fill in a questionnaire; the real number was about 10, with the same pattern for borrowing a phone or asking directions. The authors attribute it precisely to underweighting the social cost of refusal. Multiple experiments, widely replicated since (Bohns, 2016, Current Directions in Psychological Science) — hard.
• Fairly hard · channel decides it: Roghanizad & Bohns (2017, Journal of Experimental Social Psychology): the same request succeeded far more often in person than by email, while participants predicted the two would be similar. Discount: the headline multiple is a ratio from specific experimental conditions — the direction is reliable, the number shouldn't be extrapolated.
• Hard evidence · rejection genuinely hurts, but don't overstate it: Eisenberger, Lieberman & Williams (2003, Science) found social exclusion raised activity in the dorsal anterior cingulate cortex, overlapping regions that process physical pain — the fear of asking isn't precious. But discount: Woo et al. (2014, Nature Communications) used multivoxel pattern analysis to show the two kinds of pain have separable neural representations, so "social rejection is literally the same pain system" goes too far.

Boundary: this holds for small favours. The bigger the ask — a reference, a guarantee, money — the less the underestimation applies, and the answer there is to build the balance first rather than borrow courage from statistics. Also, "in person works better" is not a licence to corner people: leaning on someone's inability to refuse in the moment crosses the same persuasion-versus-manipulation line the influence-principles piece draws. The test is whether you dare write "saying no is completely fine" into the ask, and mean it.