Does Kindness Actually Spread? Three Studies, Three Different Answers

Cooperation really does cascade three degrees through a network. It also fades within two rounds — and a feed full of good news can talk you out of acting.

There is a claim you hear constantly in this corner of the internet, and we have made it ourselves: do something good and it travels. One person helps another, that person helps a third, and the thing keeps moving outward. It sounds true. The research says it is true, with conditions attached, and some of those conditions cut against the whole idea.

Start with the study that made the claim respectable. In 2010 James Fowler and Nicholas Christakis published a paper in PNAS on what they called cooperative behavior cascades. They used laboratory public goods games in which people were randomly reshuffled into new groups of strangers each round, so nobody ever played the same person twice. That design matters, because it means any spread they measured could not be simple tit-for-tat. You cannot repay someone you will never meet again.

Generosity moved anyway. Each extra contribution a person made in the first round was, in effect, tripled over the course of the experiment by other people who had been directly or indirectly influenced to give more. The influence reached three degrees of separation — your kindness touched the friend of a friend of a friend, none of whom knew where it came from.

That is the good news.

It fades fast

In 2016 a team led by Yutaka Horita at Hokkaido University ran a chained gift-giving experiment with 131 students. Each participant could donate money, doubled for whoever received it, or keep it. In one version of the game they could see what the person upstream of them had chosen. Up to 45 decisions each, across five parallel chains.

Pay-it-forward cooperation showed up. Then it evaporated. The effect was statistically significant only in the first round or two and declined steadily after that. A parallel version of the game, in which people could instead see the decisions of those downstream, where reputation was on the line, held steady across every round. Average cooperation came out at 0.42 in the pay-it-forward condition against 0.62 in the reputation condition.

The authors' reading of that is harder to sit with. Early cooperation, they concluded, came from participants who were already prosocial, not from any reciprocal impulse triggered by receiving. Prosocial individuals were roughly 7.5 times more likely to cooperate in the pay-it-forward game. The chain converted nobody. It found the people who were going to give anyway.

The part that works backwards

Milena Tsvetkova and Michael Macy ran a large online experiment, published in PLOS ONE in 2014, separating two mechanisms that usually get bundled together: receiving help makes you likelier to help someone else, and simply watching others help makes you likelier to do the same. Both hold up. But they also found something that ought to give pause to anyone running a wall of good news — observing widespread generosity can weaken your sense that your own contribution is needed.

If everyone is already helping, why you? A feed full of rescues and fundraisers can talk you out of your own small act.

What this changes for our map

We run a map where people post things they did for each other. These three studies add up to a warning about what we imagine that map is doing.

It does spread something. The cascade finding is solid and the design ruled out the easy explanations. But it spreads wide, not deep, and mostly in the first round or two. It works better when the person acting is visible to someone whose opinion they care about, which argues for names and neighbourhoods over anonymous counters. A map whose main message is how much good is already happening may quietly achieve the opposite of what it intends.

So we shouldn't hide the good news — just show the gap next to it. A request still sitting unanswered does more work than a story about one that was.

None of this is settled. The Fowler and Christakis result has been argued over for fifteen years, the Horita experiment rested on 131 Japanese undergraduates and may not travel, and online experiments have the usual problems of small, self-selected samples. The direction is consistent enough to act on, though: kindness moves, it moves less far than the slogan promises, and the surest way to stop it moving is to convince people it is already handled.

We followed the pay-it-forward thread out of the lab and into a Florida coffee queue in 378 coffees paid forward, and why generosity shrinks along the way.

Sources: Fowler JH, Christakis NA, "Cooperative behavior cascades in human social networks", PNAS 2010; 107(12): 5334–5338 (doi:10.1073/pnas.0913149107). Horita Y, Takezawa M, Kinjo T, Nakawake Y, Masuda N, "Transient nature of cooperation by pay-it-forward reciprocity", Scientific Reports 2016; 6: 19471 (doi:10.1038/srep19471). Tsvetkova M, Macy MW, "The social contagion of generosity", PLOS ONE 2014; 9(2): e87275 (doi:10.1371/journal.pone.0087275).