Back to the directoryFind in the timeline
Facebook's emotional contagion experiment
Kramer, Guillory and Hancock (2014) filtered News Feed posts for 689,003 Facebook users for one week and counted emotional words in users' own posts. Effects were very small, and PNAS published an Expression of Concern about consent.
- Where it sits
- Academic psychologyResearch published for other researchers and students.
- Evidence level
- DocumentedThe event or practice is documented by primary or authoritative sources.
- Organizations
- Facebook (Meta)An organization is listed where the sources concern it. The text says exactly how.
- Sources inspected
- 3 of 33 with the relevant text, company document or archive item checked, 0 as an abstract or reference article, 0 not verified.
What happened
In 2014 PNAS published Kramer, Guillory and Hancock's paper on emotional contagion through Facebook's News Feed. For one week, 11 to 18 January 2012, the team changed what 689,003 users saw by omitting some posts that contained positive or negative words. After public criticism over consent, PNAS published an Editorial Expression of Concern. Adam Kramer replied on his Facebook page in late June 2014.
The psychological idea
Each of two experiments, one with fewer positive posts and one with fewer negative posts, had a control in which posts were omitted at random regardless of emotion. The outcome was the percentage of positive or negative words in users' own posts, counted with the LIWC2007 word lists. The paper tests whether emotional expression spreads through a feed. It measures word use, not mood.
In practice
The documented result is small. Cohen's d was 0.02 for the main word-use changes and 0.001 for the smallest one. The authors say small effects can have large aggregated consequences at Facebook's scale. Kramer later wrote that the team did not clearly state its motivations and that the research benefits may not have justified the anxiety it caused. Facebook ran the manipulation inside its own product.
Why it mattered
The paper became a focus of criticism over consent in company experiments. The journal's own Expression of Concern recorded that questions had been raised about informed consent and the chance to opt out.
Limits and qualifications
It shows small changes in word use, not changes in mood. The authors treated acceptance of Facebook's Data Use Policy as informed consent. PNAS said the data collection may have involved practices not fully consistent with consent and opt-out principles, while noting that Facebook, as a private company, was not obliged to follow the Common Rule. Cornell's review board found the project outside its human research program because Facebook ran it. Kramer's reply is company self-report.
Evidence behind the claims
The experiment covered 689,003 Facebook users for one week, 11 to 18 January 2012, with each post containing emotional words having a 10 to 90 per cent chance of omission from a given News Feed view; control conditions omitted posts at random.
The outcome was the percentage of positive or negative words in users' own posts, classified with LIWC2007 word lists.
Reported effect sizes were Cohen's d of 0.02 for the main word-use changes and 0.001 for the smallest; the authors wrote that small effects can have large aggregated consequences at Facebook's scale.
The authors said the procedure was consistent with Facebook's Data Use Policy, which they treated as informed consent for the research.
PNAS's Editorial Expression of Concern said data collection may have involved practices not fully consistent with informed consent and opt-out principles, that Cornell's review board found the project outside its human research program because Facebook conducted it for internal purposes, and that Facebook was not obliged to follow the Common Rule.
Kramer said on his Facebook page that the team did not clearly state its motivations and that the research benefits may not have justified the anxiety; the PCWorld report is dated 30 June 2014.
People involved
- Adam D. I. Kramer. Facebook data scientist, lead author, replied publicly in 2014
- Jamie E. Guillory. Co-author, affiliated with Cornell
- Jeffrey T. Hancock. Co-author, affiliated with Cornell
In the Psy101 course
- OpenStax Psychology 2e, section 2.4: EthicsChapter 2, Psychological Research. Ethics. The section discusses this topic.The ethics section covers informed consent and institutional review boards, the issues raised here.
- OpenStax Psychology 2e, section 2.2: Approaches to ResearchChapter 2, Psychological Research. Method. The section discusses this topic.The section covers experiments as the method that supports cause and effect claims; it does not discuss effect size.
This is course relevance only. It is not evidence for the historical or corporate claims on this page.See the full course map.
Sources and inspection
3 of 3 sources were opened and read at the level shown. A source marked as not verified supports no claim on this page by itself.
- [1]Experimental evidence of massive-scale emotional contagion through social networksKramer, A; Guillory, J; Hancock, J. Proceedings of the National Academy of Sciences 111(24), 2014. Journal article. Primary source.Relevant text checkedRead the open-access paper on PubMed Central as returned by the fetch tool: sample, dates, omission design, LIWC measure, Cohen's d values, consent statement. Not every section was read.
- [2]Editorial Expression of Concern: Experimental evidence of massive-scale emotional contagion through social networksVerma, I. Proceedings of the National Academy of Sciences 111(29): 10779, 2014. Journal article. Critique or replication.Relevant text checkedRead the one-page notice on PubMed Central as returned by the fetch tool. The PNAS citation is 3 July 2014 (111(29):10779); the PMC page shows a 22 July issue date.
- [3]Facebook researcher defends controversial psychology experiment on usersPCWorld. PCWorld, 2014. News report. Secondary account.Relevant text checkedRead the news report as returned by the fetch tool; it dates itself 30 June 2014 and quotes Kramer's Facebook post. The post itself was not opened.
Access note: The PNAS journal pages returned 403 in the research stage, so the paper and notice were read on PubMed Central. Kramer's own post and Cornell's statement were not opened.
Independent source review: Claude review agent, independent of the research agent, 2026-10-21. Verdict: pass-with-edits.