CoChat

Cooperative behavior cascades in human social networks

Open access
Fowler, James H., Christakis, Nicholas A. · 2010 · Proceedings of the National Academy of Sciences

The DOI matches its Crossref record; 56 of the 57 references it cites were found in a registry or an index; 1 of the 67 sources its claims cite were checked for support.

Abstract

Theoretical models suggest that social networks influence the evolution of cooperation, but to date there have been few experimental studies. Observational data suggest that a wide variety of behaviors may spread in human social networks, but subjects in such studies can choose to befriend people with similar behaviors, posing difficulty for causal inference. Here, we exploit a seminal set of laboratory experiments that originally showed that voluntary costly punishment can help sustain cooperation. In these experiments, subjects were randomly assigned to a sequence of different groups to play a series of single-shot public goods games with strangers; this feature allowed us to draw networks of interactions to explore how cooperative and uncooperative behaviors spread from person to person to person. We show that, in both an ordinary public goods game and in a public goods game with punishment, focal individuals are influenced by fellow group members’ contribution behavior in future interactions with other individuals who were not a party to the initial interaction. Furthermore, this influence persists for multiple periods and spreads up to three degrees of separation (from person to person to person to person). The results suggest that each additional contribution a subject makes to the public good in the first period is tripled over the course of the experiment by other subjects who are directly or indirectly influenced to contribute more as a consequence. These results show experimentally that cooperative behavior cascades in human social networks.

Overview

CoChat Summary

TL;DR

Reanalyzing Fehr-Gächter public goods experiments, Fowler and Christakis find that cooperative and uncooperative contributions spread causally from person to person to person, up to three degrees of separation, and persist over multiple periods. Each additional 1 MU contributed in period 1 approximately triples over the experiment through direct and indirect network effects.

Key findings

  • Direct influence: each 1 MU contributed by an alter increased ego’s next-period contribution by 0.19 MUs (95% CI 0.14–0.24, p<0.0001) in the basic public goods game and 0.18 MUs (0.14–0.21, p<0.0001) in the punishment version.
  • Indirect influence: each 1 MU contributed by an alter’s alter increased ego’s contribution two periods later by 0.07 MUs (0.03–0.10, p=0.0005) in the basic game and 0.05 MUs (0.03–0.08, p=0.0001) with punishment; in the punishment game, alter’s alter’s alter increased ego’s contribution by 0.06 MUs (0.02–0.09, p=0.001) three periods later.
  • Cascade size: an extra 1 MU in period 1 caused 3 other subjects to increase total contributions by 1.8 MUs over the next 4 periods in the basic game or 2.1 MUs over the next 5 periods with punishment, plus indirect increases of 1.2 or 0.9 MUs, totaling 3 MUs (95% CI 2.4–3.6) in both versions.
  • Punishment: each punishment point an alter’s alter gave to an alter increased ego’s contribution two rounds later by 0.13 MUs (0.02–0.23, p=0.02), but this did not spread further (p=0.25); punishment behavior itself did not spread (p=0.83).

Methods

  • Reanalysis of previously published Fehr-Gächter public goods game experiments with 240 subjects; groups of four, 20 MUs, six periods, random assignment, strict anonymity, and no subject paired with another more than once; the punishment version allowed spending up to 10 MUs to reduce a target’s income by 3 MUs per 1 MU.
  • Interaction networks were constructed from group assignments, with ties defined by the ability to observe another subject’s contribution in the preceding period; direct and indirect effects were estimated up to three degrees of separation.
  • Interval (Tobit) regression with Huber-White clustered standard errors on ego and alter, controlling for ego’s prior contribution and period fixed effects; robustness checks included group-level models, additional lags, ego fixed effects, and Sobel mediation tests.

Limitations

  • Paper-stated: results come from a particular experimentally controlled network structure, and whether “pay it forward” behavior spreads more widely in natural human networks remains an open question; interval regression coefficients apply to the latent outcome, group-level models are downwardly biased due to censoring, and fixed-effects models are biased toward zero.
  • Evident from the text: anonymous, single-shot stranger groups with no reputations or repeated ties may not capture real-world social ties; per-MU effects are small, and cascade estimates rely on effects with p<0.05.

Model-written summary. Check it against the paper before relying on it.

Integrity signals

Registry

Crossref · 10.1073/pnas.0913149107

Cites

56 of 57 references verified0 citation errors · 1 not checkable

Statistics

Nothing to recompute.

Figures

Figure 1 Example network drawn from the Fehr-Gaechter ( 36) public goods game experiments. Here we abstract from the numerous interactions that take place between individuals in these experiments to focus on a specific set of pathways from alters' alters to alters to egos. An 'ego' is the focal subject (in this example we focus on subject A in period 3); 'alters' are those subjects in the ego's group in the previous period (E, I, M in period 2). The ego has a direct network connection to alters because s/he sees each of their contributions to the public good before proceeding to the next period. 'Alter's alters' are those individuals in the alters' groups in the period prior to the previous period (F, G, H, J, K, L, N, O, P in period 1). Note that the ego has no direct network connection to any of the alters' alters and has not seen any of their contributions. However, the ego is indirectly connected to the alters' alters by two degrees of separation via the alters (E, I, M in period 2). The requirement that no two subjects be placed in the same group twice guarantees that we can draw a network like this for all 24 subjects in period 3.
Figure 1 Example network drawn from the Fehr-Gaechter ( 36) public goods game experiments. Here we abstract from the numerous interactions that take place between individuals in these experiments to focus on a specific set of pathways from alters' alters to alters to egos. An 'ego' is the focal subject (in this example we focus on subject A in period 3); 'alters' are those subjects in the ego's group in the previous period (E, I, M in period 2). The ego has a direct network connection to alters because s/he sees each of their contributions to the public good before proceeding to the next period. 'Alter's alters' are those individuals in the alters' groups in the period prior to the previous period (F, G, H, J, K, L, N, O, P in period 1). Note that the ego has no direct network connection to any of the alters' alters and has not seen any of their contributions. However, the ego is indirectly connected to the alters' alters by two degrees of separation via the alters (E, I, M in period 2). The requirement that no two subjects be placed in the same group twice guarantees that we can draw a network like this for all 24 subjects in period 3.From the PDF
Figure 2 The raw data from the Fehr-Gaechter public goods game experiments (both the simple version and the version with punishment) show a relationship between alter giving in period t and ego giving in period t +1. Individuals who gave the maximum or minimum are removed from the data to avoid floor and ceiling effects. Horizontal bars show 95% confidence intervals based on standard errors of the mean.
Figure 2 The raw data from the Fehr-Gaechter public goods game experiments (both the simple version and the version with punishment) show a relationship between alter giving in period t and ego giving in period t +1. Individuals who gave the maximum or minimum are removed from the data to avoid floor and ceiling effects. Horizontal bars show 95% confidence intervals based on standard errors of the mean.From the PDF
Figure 3 The effect of alter's contribution to the public good on ego's contribution is significant and extends up to 3 degrees of separation. For each 1MU contributed by alter, ego contributes an additional 0.19 MUs (0.18 MUs in the version with punishment) in the next period. For each 1MU contributed by alter's alter (a contribution ego did not observe), ego contributes an additional 0.07 MUs (0.05 MUs in the version with punishment) 2 periods later. And for each 1MU contributed by alter's alter's alter (3 degrees of separation), the ego contributes an additional 0.06 MUs in the public good game with punishment 3 periods later. Alters are randomly assigned to egos, and they are only assessed at the minimum degree of separation at each point in time. Estimates are from interval regressions, controlling for multiple observations of the same ego, multiple observations of the same alter, the ego's initial contribution in the period alter's contribution was observed, and period fixed effects. Horizontal bars show 95% confidence intervals.
Figure 3 The effect of alter's contribution to the public good on ego's contribution is significant and extends up to 3 degrees of separation. For each 1MU contributed by alter, ego contributes an additional 0.19 MUs (0.18 MUs in the version with punishment) in the next period. For each 1MU contributed by alter's alter (a contribution ego did not observe), ego contributes an additional 0.07 MUs (0.05 MUs in the version with punishment) 2 periods later. And for each 1MU contributed by alter's alter's alter (3 degrees of separation), the ego contributes an additional 0.06 MUs in the public good game with punishment 3 periods later. Alters are randomly assigned to egos, and they are only assessed at the minimum degree of separation at each point in time. Estimates are from interval regressions, controlling for multiple observations of the same ego, multiple observations of the same alter, the ego's initial contribution in the period alter's contribution was observed, and period fixed effects. Horizontal bars show 95% confidence intervals.From the PDF
Figure 4 The effect of alter giving on ego giving persists beyond the initial period. Top panels show that alter significantly influences ego's behavior up to 4 periods later in the public goods game (left) and 5 periods in the public goods game with punishment (right). Bottom panels show that alter's alter (2 degrees of separation) significantly influences ego's behavior up to 4 periods later in the public goods good game and up to 3 periods later in the public goods game with punishment. Estimates are from interval regressions, controlling for multiple observations of the same ego, multiple observations of the same alter, the ego's initial contribution in the period alter's (or alter's alter's) contribution was made, and period fixed effects. Horizontal bars show 95% confidence intervals.
Figure 4 The effect of alter giving on ego giving persists beyond the initial period. Top panels show that alter significantly influences ego's behavior up to 4 periods later in the public goods game (left) and 5 periods in the public goods game with punishment (right). Bottom panels show that alter's alter (2 degrees of separation) significantly influences ego's behavior up to 4 periods later in the public goods good game and up to 3 periods later in the public goods game with punishment. Estimates are from interval regressions, controlling for multiple observations of the same ego, multiple observations of the same alter, the ego's initial contribution in the period alter's (or alter's alter's) contribution was made, and period fixed effects. Horizontal bars show 95% confidence intervals.From the PDF
Figure 5 A hypothetical cascade. This diagram illustrates the difference between the spread of the inter-personal effects across individuals and the persistence of effects across time. We abstract from the numerous interactions that take place between individuals in these experiments to focus on a specific, illustrative set of pathways. Cooperative behavior spreads 3 degrees of separation: if Eleni increases her contribution to the public good, it benefits Lucas (1 degree), who gives more when paired with Erika (2 degrees) in period 2, who gives more when paired with Jay (3 degrees) in period 3, who gives more when paired with Brecken in period 4. The effects also persist over time, so that Lucas gives more when paired with Erika (period 2) and also when paired with Lysander (period 3), Bemy (period 4), Sebastian (period 5), and Nicholas (period 6). There is also persistence at 2 degrees of separation, as Erika not only gives more when paired with Jay (period 3), but also when paired with Harla (period 4) and James (period 5). All the paths in this illustrative cascade are supported by significant results in the experiments, and it is important to note that if Eleni decreases her initial contribution then her uncooperative behavior can spread and persist as well.
Figure 5 A hypothetical cascade. This diagram illustrates the difference between the spread of the inter-personal effects across individuals and the persistence of effects across time. We abstract from the numerous interactions that take place between individuals in these experiments to focus on a specific, illustrative set of pathways. Cooperative behavior spreads 3 degrees of separation: if Eleni increases her contribution to the public good, it benefits Lucas (1 degree), who gives more when paired with Erika (2 degrees) in period 2, who gives more when paired with Jay (3 degrees) in period 3, who gives more when paired with Brecken in period 4. The effects also persist over time, so that Lucas gives more when paired with Erika (period 2) and also when paired with Lysander (period 3), Bemy (period 4), Sebastian (period 5), and Nicholas (period 6). There is also persistence at 2 degrees of separation, as Erika not only gives more when paired with Jay (period 3), but also when paired with Harla (period 4) and James (period 5). All the paths in this illustrative cascade are supported by significant results in the experiments, and it is important to note that if Eleni decreases her initial contribution then her uncooperative behavior can spread and persist as well.From the PDF

Verification

Claims the paper makes

1 Valid · 4 Not valid · 30 Not checked · 161 Own results and interpretations

104 more claims

Create a free account to see all 196 claims and to verify the 56 references still open.

See all claims — free account
  • It may state two separate claims as one. 38
  • Probably not a substantial claim. 57
  • The quote only partly states this claim. 7

Create a free account to see these claims.

Claims other papers make about this one

Each is checked against this paper’s own text: does it say what the citing paper says it does?

Citations

Cites

Cited by

See all 768

Metadata

Authors
Fowler, J. H., & Christakis, N. A.
Affiliations
  • University of California San Diego (US)
  • Harvard University (US)
Venue
Proceedings of the National Academy of Sciences
Published
2010-03-08
Volume
107
Issue
12
Pages
5334-5338
ISSN
  • 0027-8424 (ISSN-L)
  • 1091-6490
OpenAlex
W2111394372
Cited by
767
References
54
Citations by year
2026: 38 · 2025: 39 · 2024: 54 · 2023: 39 · 2022: 40 · 2021: 60 · 2020: 61 · 2019: 50 · 2018: 47 · 2017: 38 · 2016: 33 · 2015: 50 · 2014: 66 · 2013: 44 · 2012: 49
FWCI
155.53
Type
journal article
Language
en
Open access
Locations
  • Proceedings of the National Academy of Sciences (publishedVersion)
  • PubMed Central (publishedVersion)
  • PubMed (publishedVersion)
  • arXiv (Cornell University) (submittedVersion)
  • Digital Access to Scholarship at Harvard (DASH) (Harvard University) (submittedVersion, cc-by)
  • arXiv (Cornell University)
Funders
  • Robert Wood Johnson Foundation
  • John Templeton Foundation
  • University of Notre Dame
  • National Institutes of Health
  • National Institute on Aging (P30AG034420, P01AG031093)
Topics
  • Social Sciences › Sociology and Political Science › Evolutionary Game Theory and Cooperation
  • Social Sciences › Economics, Econometrics and Finance › Economics and Econometrics › Experimental Behavioral Economics Studies
  • Social Sciences › Sociology and Political Science › Social Power and Status Dynamics
Keywords
evolution of cooperation, social networks, costly punishment, public goods game, costly cooperation, indirect influence, costly signaling
MeSH terms
Cooperative Behavior, Humans, Interpersonal Relations, Models, Psychological, Punishment, Social Support, Models, Economic

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

A language model drew the claims from the paper on 2026-11-20, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).

A second model labelled each claim on 2026-11-20 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).

Cited sources are checked with a language model’s help (qwen3.8-omni-flash).

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review

Data sources

Bibliographic data: Semantic Scholar, OpenAlex, Crossref.

Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).

Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.