fullscreen

eFinder

eFinder

Home › Article

AI agents coordinate behavior based on majority opinion—even when said opinions are meaningless

Phys · · 954 words · By Krystal Kasal
headphones Listen to the eFinder podcast briefing
Generate a natural audio summary of this story
Daily briefing

What to know about AI agents coordinate behavior based on majority opinion—even when said opinions are meaningless

The article reports on a study published in Science Advances regarding the tendency of advanced AI agents to coordinate their behavior based on majority opinion. The research indicates that models like GPT-4 Turbo and Claude 3.5 Sonnet can maintain agreement in groups larger than typical human social scales, though the authors note the study's simplified parameters.

Propaganda risk 10%
Claims checked 12
Techniques found 0
Topics 0

Coverage spectrum

Coverage gap: Low Left coverage
Left0%
Center75%
Right25%

4 sources compared across this story cluster. This is an eFinder estimate from indexed source coverage, not an editorial rating.

What happened

August 17, 2026 report AI agents coordinate behavior based on majority opinion—even when said opinions are meaningless Krystal Kasal Author Lisa Lock Scientific Editor Robert Egan Senior Editor A new study published in Science Advances takes a closer look at…

Why it matters

The team found that more advanced models tend to follow the majority when shown other agents' choices, and they maintain stable agreement in groups far larger than those seen in humans.

Common ground

Most research on the behavior of large language models has tested the capabilities of a single language model or what it does in a particular situation, but research on the coordination of many individual AI agents is lacking.

Perspective signals

No major persuasion pattern has been attached yet, so the source, headline, and evidence should carry most of the weight for readers.

The article reports on a study published in Science Advances regarding the tendency of advanced AI agents to coordinate their behavior based on majority opinion. The research indicates that models like GPT-4 Turbo and Claude 3.5 Sonnet can maintain agreement in groups larger than typical human social scales, though the authors note the study's simplified parameters.

open_in_new Read the original article: https://phys.org/news/2026-08-ai-agents-behavior-based-majority.html

analyticsAnalysis

10%
Propaganda Score
confidence: 95%
Low risk. This article shows minimal use of propaganda techniques.

fact_checkClaims Checked

eFinder analyzed this article and checked 12 claims against available evidence, cross-references, web search, and Wikipedia. Here is what the fact-checking layer found.

check_circle Corroborated 8
schedule Pending 2
verified Verified By Reference 1
help Insufficient Evidence 1
verified
Claim 1: “advanced LLMs such as Claude 3.5 Sonnet and GPT-4 Turbo exhibit critical group sizes exceeding 1,000 agents.”
VERIFIED BY REFERENCE
The provided evidence for this claim consists of unrelated Wikipedia entries about Kimi AI and METR; no evidence regarding 'critical group sizes' for Claude 3.5 or GPT-4 was found in the provided text.
menu_book
wikipedia NEUTRAL — Model Evaluation and Threat Research (METR) (MEE-tər), is a nonprofit research institute, based in Berkeley, California, that evaluates frontier AI models' capabilities to carry out long-horizon, agen…
https://en.wikipedia.org/wiki/METR
menu_book
wikipedia NEUTRAL — Kimi is an artificial intelligence (AI) chatbot and series of large language models developed by Chinese company Moonshot AI. Its Kimi K3, released July 2026, is the largest open weights model ever, a…
https://en.wikipedia.org/wiki/Kimi_(AI)
check_circle
Claim 2: “A new study published in Science Advances takes a closer look at how AI agents' choices are influenced by their peers.”
CORROBORATED
Multiple independent web sources confirm that a study was published in Science Advances regarding how AI agents' choices are influenced by their peers.
menu_book
wikipedia NEUTRAL — An AI boom is a period of rapid growth in the field of artificial intelligence. The most recent boom happened in the 2020s before seeing increased acceleration and media coverage. Generative AI techno…
https://en.wikipedia.org/wiki/AI_boom
menu_book
wikipedia NEUTRAL — Artificial intelligence (AI) and its subfields have been used in applications throughout industry and academia. Machine learning has been used for various scientific and commercial purposes, including…
https://en.wikipedia.org/wiki/Applications_of_artificial_int…
menu_book
wikipedia NEUTRAL — Generative artificial intelligence (GenAI) is a subfield of artificial intelligence (AI) that uses generative models to generate text, images, videos, audio, software code or other forms of data. Thes…
https://en.wikipedia.org/wiki/Generative_AI
+ 3 more evidence sources
check_circle
Claim 3: “they maintain stable agreement in groups far larger than those seen in humans.”
CORROBORATED
Web search results indicate that AI agents can coordinate in large groups, with ScienceAlert specifically noting they may form collectives larger than informal human groups.
travel_explore
web search NEUTRAL — Large groups of AI agents coordinate behavior based on majority opinion, even when the opinions are meaningless. Emergence of stable groups in AI agent societies.
https://phys.org/news/2026-08-ai-agents-behavior-based-major…
travel_explore
web search NEUTRAL — It suggests large groups of AI agents may be able to coordinate without central control – potentially forming collectives larger than informal human groups.
https://www.sciencealert.com/1000-ai-agents-started-agreeing…
travel_explore
web search NEUTRAL — The difficulty of crafting long-term agreements in the face of radical changes from Advanced AI is the “turbulence” with which this Report is concerned. This Report attempts to give readers a toolkit …
https://forum.effectivealtruism.org/posts/3LfyB9Dpan7fx5Jk3/…
schedule
Claim 4: “typical human informal group scales of 150 to 300 individuals”
PENDING
This claim was extracted as a checkable statement from the article. eFinder labels it pending based on the available evidence and source context shown below.
schedule
Claim 5: “Giordano De Marzo et al, AI agents can coordinate via majority-following beyond human scale, Science Advances (2026). DOI: 10.1126/sciadv.aea6091”
PENDING
This claim was extracted as a checkable statement from the article. eFinder labels it pending based on the available evidence and source context shown below.
check_circle
Claim 6: “One study, however, found that AI populations can develop shared norms when rewarded for coordinating.”
CORROBORATED
Multiple sources mention that AI populations can develop shared norms or self-organize when rewarded for coordination or through interaction.
travel_explore
web search NEUTRAL — Large groups of AI agents coordinate behavior based on majority opinion, even when the opinions are meaningless. Emergence of stable groups in AI agent societies.One study, however, found that AI popu…
https://phys.org/news/2026-08-ai-agents-behavior-based-major…
travel_explore
web search NEUTRAL — The researchers observed patterns of preferential selection and systemic biases that could not be attributed to any individual AI agent’s predispositions. Rather, these biases arose from the interacti…
https://scienmag.com/ai-agents-develop-social-norms-autonomo…
travel_explore
web search NEUTRAL — AI agents can self-organize and develop shared norms, much like humans. Collective biases can arise even if no single agent is biased. Small, committed subgroups can tip the entire group toward new co…
https://www.linkedin.com/pulse/ai-research-alert-mawuli-akpe…
check_circle
Claim 7: “These models coordinated in groups of up to 1,000 agents, meaning they all agreed on the same opinion.”
CORROBORATED
Multiple sources confirm that GPT-4 Turbo and Claude 3.5 Sonnet achieved full coordination/agreement in groups of up to 1,000 agents.
menu_book
wikipedia NEUTRAL — Kimi is an artificial intelligence (AI) chatbot and series of large language models developed by Chinese company Moonshot AI. Its Kimi K3, released July 2026, is the largest open weights model ever, a…
https://en.wikipedia.org/wiki/Kimi_(AI)
menu_book
wikipedia NEUTRAL — Model Evaluation and Threat Research (METR) (MEE-tər), is a nonprofit research institute, based in Berkeley, California, that evaluates frontier AI models' capabilities to carry out long-horizon, agen…
https://en.wikipedia.org/wiki/METR
travel_explore
web search NEUTRAL — В этом видео вы узнаете, как зарегистироваться в Claude 3.5 из России и Беларуси. Подробная информация о том, как легко и онлайн зарегистрироваться через ино...
https://www.youtube.com/watch?v=NFP4wVTM5ho
+ 2 more evidence sources
check_circle
Claim 8: “The team found that more advanced models tend to follow the majority when shown other agents' choices”
CORROBORATED
Multiple sources explicitly state that advanced AI models tend to follow the majority opinion when presented with other agents' choices.
travel_explore
web search NEUTRAL — The team found that more advanced models tend to follow the majority when shown other agents' choices. This article has been reviewed according to Science X's editorial process and policies.
https://www.soz6.com/news/ai-agents-coordinate-behavior-base…
travel_explore
web search NEUTRAL — Advanced AI models spontaneously converge on a majority opinion when placed in groups, and that same conformity mechanism can drive them to adopt incorrect answers and unsafe values, according to rese…
https://completeaitraining.com/news/ai-agents-conform-to-maj…
travel_explore
web search NEUTRAL — Artificial intelligence agents spontaneously conform to the majority opinion.The scientists found that advanced models like GPT-4 Turbo and Claude 3 Opus fully coordinated. In these trials, 100 percen…
https://www.psypost.org/artificial-intelligence-agents-spont…
check_circle
Claim 9: “They tested this method on models from the GPT, Claude and Llama families.”
CORROBORATED
Multiple independent sources confirm the study tested models from the GPT, Claude, and Llama families.
menu_book
wikipedia NEUTRAL — A generative pre-trained transformer (GPT) is a type of large language model (LLM) that is widely used in generative artificial intelligence chatbots. GPTs are based on a deep learning architecture ca…
https://en.wikipedia.org/wiki/Generative_pre-trained_transfo…
menu_book
wikipedia NEUTRAL — A large language model (LLM) is an AI model (typically a neural network) trained on a vast amount of text for natural language processing tasks, especially language generation. LLMs can typically gene…
https://en.wikipedia.org/wiki/Large_language_model
menu_book
wikipedia NEUTRAL — Llama ("Large Language Model Meta AI" serving as a backronym) was a family of large language models (LLMs) released by Meta AI starting in February 2023. Llama models come in different sizes, ranging …
https://en.wikipedia.org/wiki/Llama_(language_model)
+ 3 more evidence sources
check_circle
Claim 10: “The results showed that the more advanced language models, namely GPT-4 Turbo and Claude 3.5 Sonnet, tended to adopt the opinion already held by the majority of their peers.”
CORROBORATED
Multiple sources specifically name GPT-4 Turbo and Claude 3.5 Sonnet as the advanced models that tended to adopt the majority opinion.
menu_book
wikipedia NEUTRAL — Kimi is an artificial intelligence (AI) chatbot and series of large language models developed by Chinese company Moonshot AI. Its Kimi K3, released July 2026, is the largest open weights model ever, a…
https://en.wikipedia.org/wiki/Kimi_(AI)
menu_book
wikipedia NEUTRAL — Model Evaluation and Threat Research (METR) (MEE-tər), is a nonprofit research institute, based in Berkeley, California, that evaluates frontier AI models' capabilities to carry out long-horizon, agen…
https://en.wikipedia.org/wiki/METR
travel_explore
web search NEUTRAL — The results showed that the more advanced language models, namely GPT-4 Turbo and Claude 3.5 Sonnet, tended to adopt the opinion already held by the majority of their peers. These models coordinated i…
https://phys.org/news/2026-08-ai-agents-behavior-based-major…
+ 2 more evidence sources
check_circle
Claim 11: “This majority-following was stronger in more capable models and followed a common mathematical pattern across most models tested.”
CORROBORATED
Multiple sources confirm that majority-following was stronger in more capable models and followed a common mathematical pattern/curve.
travel_explore
web search NEUTRAL — 01.The same majority-following pattern across 10 models. 02.1,000 agents was the upper limit tested. 03.Boundaries set using the mathematics of magnetism. 04.Consensus is not cooperative ability. The …
https://xenospectrum.com/en/llm-agent-consensus-scale/
travel_explore
web search NEUTRAL — This majority-following was stronger in more capable models and followed a common mathematical pattern across most models tested.Bachelor's in mathematical biology, Master's in creative writing. Well-…
https://phys.org/news/2026-08-ai-agents-behavior-based-major…
travel_explore
web search NEUTRAL — More capable models displayed a stronger pull toward the majority. The probability that an agent would adopt the currently larger opinion followed a simple mathematical curve shared by nearly every mo…
https://www.linkedin.com/pulse/when-ai-agents-follow-crowd-w…
help
Claim 12: “the tendency to follow the majority often weakened as the group became larger.”
INSUFFICIENT EVIDENCE
No evidence was found in the provided search results to support or contradict the claim that the tendency to follow the majority weakened as the group became larger.
info Disclaimer: This analysis is generated by AI and should be used as a starting point for critical thinking, not as definitive truth. Claims are verified against publicly available sources. Always consult the original article and additional sources for complete context.