CoChat
Sign Up
Cited by 132Language and cultural evolutionCybernetics and Technology in SocietyEthics and Social Impacts of AI

Machine culture

Open access
Brinkmann, Levin, Baumann, Fabian, Bonnefon, Jean-François, Derex, Maxime, Müller, Thomas F., Nussberger, Anne-Marie, Czaplicka, Agnieszka, Acerbi, Alberto +7 more·2023·Nature Human Behaviour
DOI 10.1038/s41562-023-01742-2
Open this paper in CoChat — free accountOpen PDF

Abstract

The ability of humans to create and disseminate culture is often credited as the single most important factor of our success as a species. In this Perspective, we explore the notion of ‘machine culture’, culture mediated or generated by machines. We argue that intelligent machines simultaneously transform the cultural evolutionary processes of variation, transmission and selection. Recommender algorithms are altering social learning dynamics. Chatbots are forming a new mode of cultural transmission, serving as cultural models. Furthermore, intelligent machines are evolving as contributors in generating cultural traits—from game strategies and visual art to scientific results. We provide a conceptual framework for studying the present and anticipated future impact of machines on cultural evolution, and present a research agenda for the study of machine culture. Artificial intelligence tools and systems are increasingly influencing human culture. Brinkmann et al. argue that these ‘intelligent machines’ are transforming the fundamental processes of cultural evolution: variation, transmission and selection.

AI snapshot

Written from the paper’s full text.

TL;DR

This Perspective argues that intelligent machines are transforming cultural evolution by altering variation, transmission, and selection, producing “machine culture”—culture mediated or generated by machines. It offers a conceptual framework and research agenda for studying how generative AI, reinforcement learning, LLMs, and recommender systems reshape cultural dynamics.

Key findings

  • Machines affect all three Darwinian properties of culture: variation (generative AI recombination and reinforcement-learning exploration), transmission (LLMs as cultural models and reservoirs of knowledge), and selection (recommender systems curating content and social ties).
  • AlphaGo’s move 37 was estimated by AlphaGo itself to have a 1 in 10,000 chance of being made by a human; AlphaGo Zero played 4.9 million self-play games in three days; human Go decision quality showed little improvement from 1950 to 2016, followed by a sudden improvement after AlphaGo in March 2016.
  • LLMs such as ChatGPT act as a new mode of cultural transmission and social learning from machines; a survey projects that around 20% of the workforce will experience LLMs impacting at least half of their tasks, though this is described as an early and uncertain estimate.
  • Recommender systems do not add cultural variation but impact selective retention and transmission by rewiring social networks and filtering content; link recommendation can fuel “the rich get richer” dynamics, and engagement optimization may amplify emotionally and morally charged content, echo chambers, and filter bubbles.

Methods

  • Perspective/conceptual synthesis, not an empirical study; it defines culture, cultural evolution, social learning, AI, and machines, then maps machine impacts onto variation, transmission, and selection.
  • Draws on published examples and studies—AlphaGo, generative AI, LLMs, recommender systems—and presents Table 1 as “tentative conjectures” plus grand challenges and open questions.
  • No new data collection or formal analysis is reported; the conclusion was created by GPT-4 from a GPT-4 summary of the manuscript with minimal editing.

Limitations

  • Paper states: the projection that around 20% of the workforce will be affected is an early estimate and inherently uncertain; measuring how much of human cultural dynamics is attributable to algorithms is difficult because of confounds and feedback loops; claims about machine-generated art developing its own unique style cannot yet be reliably verified; evidence for model collapse is preliminary.
  • Text makes evident: as a Perspective, it is selective and conceptual rather than a systematic review or primary empirical test, and many claims are anticipatory or conjectural, so they may not generalize.

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

Integrity signals

Registry

Crossref · 10.1038/s41562-023-01742-2

Cites

171 of 171 references verified0 citation errors · 0 not checkable

Statistics

Nothing to recompute.

Claims the paper makes

0 of 136 cited sources checked · 0 supported · 0 flagged

  • Findingcitation

    AlphaGo managed to surprise Sedol with distinctively non-human gameplay.

    “Remarkably, AlphaGo managed to surprise Sedol with distinctively non-human gameplay.”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites [29]

      Not checked in the free run.

  • Findingcitation

    The innovations generated by AlphaGo and AlphaGo Zero soon entered human culture, as shown by research comparing human gameplay before and after the algorithms' introduction.

    “The innovations generated by AlphaGo and AlphaGo Zero soon entered human culture, as shown by research comparing human gameplay before and after the algorithms' introduction 31 .”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites [31] source

      Not checked in the free run.

  • Interpretationcitation

    Generative machines, such as text-to-image algorithms, are contributing to the variety of cultural traits.

    “Generative machines, such as text-to-image algorithms, are contributing to the variety of cultural traits.”
    Quote in the paperSubstantial claim0 of 3 sources checked
    • cites [19]

      Not checked in the free run.

    • cites [20] source

      Not checked in the free run.

    • cites [21] source

      Not checked in the free run.

  • Interpretationnone given

    Intelligent machines will exert a transformative influence on cultural evolution through their impact on all three Darwinian properties: variation, transmission and selection.

    “in this Perspective we argue that intelligent machines will exert a transformative influence on cultural evolution through their impact on all three Darwinian properties of culture: variation, transmission and selection (Fig. 1).”
    Quote in the paperNot a substantial claim
    • Rests on an assumption: The three Darwinian properties (variation, transmission, selection) are sufficient to capture the ways machines influence cultural evolution.
  • Interpretationreasoning

    Intelligent machines may increase the variety of cultural traits by building on human knowledge while identifying and deliberately avoiding common human pathways.

    “Another way in which intelligent machines may increase the variety of cultural traits is by building on human knowledge while at the same time identifying and deliberately avoiding common human pathways.”
    Quote in the paperSubstantial claim
  • May combine two claims. 44
  • Classified as probably not a substantial claim. 54

  • May combine two claims.

    Findingcitation

    AlphaGo Zero, starting from a blank slate, not only rediscovered human Go strategies but also developed strategies that surpassed those of its human creators.

    “Starting from a blank slate, it not only rediscovered human Go strategies but also developed strategies that surpassed those of its human creators.”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites [30] source

      Not checked in the free run.

  • May combine two claims.

    Findingcitation

    Human decision quality in Go showed very little improvement from 1950 to 2016, followed by a sudden improvement after the introduction of AlphaGo in March 2016.

    “The decision quality, as measured by an open-source variant of AlphaGo Zero, showed very little improvement in human gameplay from 1950 to 2016, followed by a sudden improvement after the introduction of AlphaGo in March 2016 (refs.32,33) (Fig. 3b).”
    Quote in the paperSubstantial claim0 of 2 sources checked
    • cites [32] source

      Not checked in the free run.

    • cites [33]

      Not checked in the free run.

  • May combine two claims.

    Premisecitation

    Culture exhibits variation, transmission and selection, and evolves through selective retention of cultural traits and non-selective processes such as drift.

    “Culture has been shown to exhibit variation, transmission and selection, and it evolves through the selective retention of cultural traits as well as non-selective processes such as drift 10 .”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites [10]

      Not checked in the free run.

  • May combine two claims.

    Interpretationreasoning

    We are observing not only a transformation of human culture but also its evolution into machine culture-culture mediated or generated by machines.

    “We are observing not only a transformation of human culture but also its evolution into machine culture-culture mediated or generated by machines.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Interpretationcitation

    LLMs act as models of human culture, facilitating cultural transmission across individuals and generations.

    “Trained on extensive human text data, both historical and contemporary, LLMs act as models of human culture 25 , facilitating cultural transmission across individuals and generations.”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites [25] source

      Not checked in the free run.

  • May combine two claims.

    Interpretationcitation

    Large language models are facilitating the transmission of cultural knowledge and redefining the value of human intellectual skills.

    “Large language models (LLMs) are facilitating the transmission of cultural knowledge and redefining the value of human intellectual skills.”
    Quote in the paperSubstantial claim0 of 2 sources checked
    • cites [25] source

      Not checked in the free run.

    • cites [39] source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S2 (checkable), S3 (load-bearing).

    Interpretationcitation

    Human cultural exploration, in science as in arts, will increasingly be in collaboration with machines.

    “Undoubtedly, human cultural exploration, in science as in arts, will increasingly be in collaboration with machines 78,79 .”
    Quote in the paperNot a substantial claim0 of 2 sources checked
    • cites [78] source

      Not checked in the free run.

    • cites [79] source

      Not checked in the free run.

    • Rests on an assumption: Current trends in AI development will continue.

  • May combine two claims.

    Interpretationcitation

    As LLM capabilities develop, the value of certain human skills will shift, with some skills losing value quickly, especially in language-related and cognitively demanding occupations such as translation, copywriting or proofreading.

    “Some skills may lose value quickly, especially in language-related and cognitively demanding occupations such as translation, copywriting or proofreading 39 .”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [39] source

      Not checked in the free run.

    • Rests on an assumption: Current trends in AI development will continue.

  • May combine two claims.

    Interpretationcitation

    The degree to which machines can learn from human cultural repertoires far exceeds what can be achieved by a single human being, allowing individual language models to cover a broad spectrum of languages comprehensively.

    “While size does not guarantee diversity 69 , the degree to which machines can learn from human cultural repertoires far exceeds what can be achieved by a single human being, allowing, for instance, individual language models to cover a broad spectrum of languages comprehensively 70 .”
    Quote in the paperSubstantial claim0 of 2 sources checked
    • cites [69] source

      Not checked in the free run.

    • cites [70]

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S2 (checkable), S3 (load-bearing).

    Premisecitation

    The ability of humans to create and disseminate culture is considered the single most important factor in our species' dominance on Earth.

    “The ability of humans to create and disseminate culture is considered the single most important factor in our species' dominance on Earth 1 .”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [1] source

      Not checked in the free run.

  • May combine two claims.

    Interpretationreasoning

    Machines can produce traits distinct from those produced by humans, thus potentially steering culture towards new paths.

    “However, machines, leveraging their unique capacities, can produce traits distinct from those produced by humans, thus potentially steering culture towards new paths.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Premisecitation

    AlphaFold has achieved remarkable precision in modelling intramolecular relations, enabling prediction of a protein's three-dimensional structure solely from its DNA sequence.

    “Surpassing the capabilities of any human-conceived model, the neural-network-based algorithm Alpha-Fold has achieved remarkable precision in modelling intramolecular relations, enabling it to tackle one of molecular biology's greatest challenges: predicting a protein's three-dimensional structure solely on the basis of its DNA sequence 65 .”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [65] source

      Not checked in the free run.

  • May combine two claims.

    Interpretationreasoning

    Machines have the ability to learn individually at an unparalleled scale, enabling discovery of novel cultural traits through extensive exploration.

    “Machines have the ability to learn individually at an unparalleled scale, enabling the discovery of novel cultural traits through extensive exploration.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S4 (based).

    Interpretationnone given

    AI systems have the potential to generate culturally alien traits.

    “As a result, AI systems have the potential to generate culturally alien traits.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Interpretationreasoning

    AlphaGo served as an early, quantifiable exemplar of machine culture, generating novel cultural variations through genuine, non-human innovation.

    “In summary, AlphaGo served as an early, quantifiable exemplar of machine culture, generating novel cultural variations through genuine, non-human innovation.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Interpretationcitation

    The mere recombination of cultural traits can lead to innovation and fuel cultural evolution.

    “Irrespective of whether LLMs truly exhibit understanding 71 , the mere recombination of cultural traits can lead to innovation and fuel cultural evolution 69,72 .”
    Quote in the paperSubstantial claim0 of 2 sources checked
    • cites [69] source

      Not checked in the free run.

    • cites [72] source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S2 (checkable).

    Premisecitation

    Culture has been defined as information capable of affecting individuals' behaviours that is acquired from other individuals via social transmission.

    “Culture has been defined as information capable of affecting individuals' behaviours that is acquired from other individuals via social transmission 50 .”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [50] source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S4 (based). May combine two claims.

    Interpretationnone given

    Recommender systems do not add variation in cultural traits, but they demonstrably impact the selective retention and transmission of cultural traits.

    “While recommender systems do not add variation in cultural traits, they demonstrably impact the selective retention and transmission of cultural traits.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S4 (based). May combine two claims.

    Interpretationnone given

    Machine technologies will affect each of the three properties (variation, transmission, selection), and these changes are likely to have transformative impacts on cultural evolution.

    “We anticipate that machine technologies will affect each of these three properties (Fig. 1), and these changes are likely to have transformative impacts on cultural evolution.”
    Quote in the paperNot a substantial claim
    • Rests on an assumption: The three Darwinian properties (variation, transmission, selection) are sufficient to capture the ways machines influence cultural evolution.

  • May combine two claims.

    Premisecitation

    General-purpose neural networks are universal approximators, allowing functional relationships to be learned.

    “Instead, general-purpose neural networks are universal approximators, allowing functional relationships to be learned 62,63 .”
    Quote in the paperSubstantial claim0 of 2 sources checked
    • cites [62] source

      Not checked in the free run.

    • cites [63] source

      Not checked in the free run.

  • May combine two claims.

    Interpretationreasoning

    Cultural transmission can be impacted not only by the replication of biases but also by disparities between human and algorithmic biases.

    “Cultural transmission can be impacted not only by the replication of biases but also by disparities between human and algorithmic biases.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Premisecitation

    This can restrict the spectrum of solutions that a population can derive, and even highly effective solutions may not be sustained if they conflict with pre-existing biases.

    “This can restrict the spectrum of solutions that a population can derive, and even highly effective solutions may not be sustained if they conflict with pre-existing biases 106 .”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites [106] source

      Not checked in the free run.

  • Classified as probably not a substantial claim.

    Interpretationreasoning

    The increased cognitive diversity-for instance, with regard to biases-within human-machine societies has the potential to expand the collective capacity to maintain diverse cultural artefacts.

    “As such, the increased cognitive diversity-for instance, with regard to biases-within human-machine societies has the potential to expand the collective capacity to maintain diverse cultural artefacts.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    Numerous strategies exist to address bias and fairness in machine learning models.

    “Numerous strategies exist to address bias and fairness in machine learning models 92 .”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [92] source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

    Interpretationnone given

    Intelligent machines will increasingly be involved in the preservation and transmission of cultural information.

    “Intelligent machines will increasingly be involved in the preservation and transmission of cultural information.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Interpretationcitation

    Machines can learn to seek and transmit information and can act as conversational and pedagogical agents, similar to teachers.

    “Besides serving as a persistent medium of cultural storage analogous to a book, machines can learn to seek and transmit information 79,81 and can act as conversational and pedagogical agents, similar to teachers 82 .”
    Quote in the paperSubstantial claim0 of 3 sources checked
    • cites [79] source

      Not checked in the free run.

    • cites [81] source

      Not checked in the free run.

    • cites [82] source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S2 (checkable), S3 (load-bearing), S4 (based).

    Interpretationnone given

    LLMs, with their capacity to handle vast quantities of training data, can contribute to the preservation and even enhancement of cultural diversity, provided that the training data are carefully curated with attention to diversity and representation.

    “LLMs, with their capacity to handle vast quantities of training data, can contribute to the preservation and even enhancement of cultural diversity, provided that the training data are carefully curated with attention to diversity and representation.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Premisecitation

    Humans and machines operate under different cognitive constraints and inhabit distinct socio-environmental contexts.

    “Importantly, humans and machines operate under different cognitive constraints 57 and inhabit distinct socio-environmental contexts.”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites [57] source

      Not checked in the free run.

  • May combine two claims.

    Interpretationnone given

    Users also react to recommender systems, both directly in terms of clicking, viewing and purchasing and in terms of what they produce, as content creators anticipate what will receive the widest distribution.

    “Users also react to recommender systems, both directly in terms of clicking, viewing and purchasing and in terms of what they produce, as content creators anticipate what will receive the widest distribution.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Premisecitation

    In social networks, maximizing user engagement may be achieved by promoting content congruent with users' past engagement or ingroup attitudes, or content that humans inherently attend to, such as emotionally and morally charged content.

    “In social networks, this may be achieved by promoting content congruent with users' past engagement or ingroup attitudes 134 , or content that humans inherently attend to, such as emotionally and morally charged content 135,136 .”
    Quote in the paperSubstantial claim0 of 3 sources checked
    • cites [134] source

      Not checked in the free run.

    • cites [135] source

      Not checked in the free run.

    • cites [136] source

      Not checked in the free run.

  • May combine two claims.

    Interpretationcitation

    Symbolic representations may remain crucial for efficient computation, but algorithmic-assisted discovery could lessen the need for their transmission, as these representations could be easily regenerated.

    “While symbolic representations may remain crucial for efficient computation, algorithmic-assisted discovery could lessen the need for their transmission, as these representations could be easily regenerated 113 .”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [113] source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S2 (checkable), S3 (load-bearing), S4 (based). May combine two claims.

    Interpretationnone given

    While recommender systems have so far been mostly shaping user preferences implicitly, LLMs may accelerate developments where users are increasingly persuaded explicitly through interactive argumentation.

    “While recommender systems have so far been mostly shaping user preferences implicitly (for example, by optimizing the position or ranking of content), LLMs may accelerate developments where users are increasingly persuaded explicitly through interactive argumentation.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Interpretationnone given

    While machines have most commonly relied on exploiting explicit user preferences and historical behaviour, there has been a growing interest in considering users' internal states to improve algorithmic recommendations.

    “While machines have most commonly relied on exploiting explicit user preferences and historical behaviour (for example, ratings and engagement), there has been a growing interest in considering users' internal states to improve algorithmic recommendations.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Premisecitation

    User engagement is a signal of value to both users and platforms deploying algorithms, connecting them in complex feedback loops.

    “It is important to note that user engagement is a signal of value to both users and platforms deploying algorithms, connecting them in complex feedback loops 144,145 :”
    Quote in the paperSubstantial claim0 of 2 sources checked
    • cites [144]

      Not checked in the free run.

    • cites [145] source

      Not checked in the free run.

  • May combine two claims.

    Premisecitation

    X's 'who to follow' recommendation was observed to disproportionately benefit those users who were already the most popular, fuelling 'the rich get richer' dynamics.

    “X's”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites [122] source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    Recommender systems might account for the cognitive cost of exploring items or learning about them by prioritizing items that are easier for users to evaluate or learn.

    “As another example, recommender systems might account for the cognitive cost of exploring items or learning about them by prioritizing items that are easier for users to evaluate or learn 128 .”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [128] source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

    Premisenone given

    Content-based filtering algorithms aim to maximize the similarity between items a user previously showed interest in and unobserved items.

    “For instance, content-based filtering algorithms aim to maximize the similarity between items a user previously showed interest in and unobserved items.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S4 (based). May combine two claims.

    Interpretationnone given

    In the context of machine culture, selection can occur at a level where machines select what and from whom humans learn, at a level where humans select machine behaviour, and at a level where there is selection between humans and machines.

    “In the context of machine culture, selection can occur at a level where machines select what and from whom humans learn, at a level where humans select machine behaviour, and at a level where there is selection between humans and machines.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Premisecitation

    Bayesian approaches can model users' uncertainty and be used to update algorithmic recommendations as the user interacts with the system.

    “Bayesian approaches can model users' uncertainty and be used to update algorithmic recommendations as the user interacts with the system 127 .”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [127] source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S2 (checkable), S4 (based).

    Interpretationreasoning

    Theories may retain importance in shaping human understanding and intuition, serving as essential tools for conceptualizing knowledge.

    “However, theories may retain importance in shaping human understanding and intuition, serving as essential tools for conceptualizing knowledge-a function that the framework we present in this work aims to fulfil.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S2 (checkable), S3 (load-bearing), S4 (based).

    Interpretationreasoning

    With the advent of AlphaFold, scientists might focus on collecting and preserving further ground-truth data on molecule structures to refine future models rather than refining and transmitting theories of atomic interactions.

    “For instance, with the advent of AlphaFold, scientists might focus on collecting and preserving further ground-truth data on molecule structures to refine future models rather than refining and transmitting theories of atomic interactions.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Premisereasoning

    Compression processes have an important role in human transmission to mitigate cultural loss: raw information often is not learnable because it is too complex, whereas a compressed rule might very well be learnable.

    “Compression processes have an important role in human transmission to mitigate cultural loss: raw information often is not learnable because it is too complex, whereas a compressed rule might very well be learnable.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Premisecitation

    Affective recommender systems use techniques such as natural language processing to make inferences about users' emotional states and even combine them with other context information such as the recommendation domain.

    “Affective recommender systems 129 use techniques such as natural language processing to make inferences about users' emotional states and even combine them with other context information such as the recommendation domain (for example, music or movies) [130][131][132] .”
    Quote in the paperNot a substantial claim0 of 3 sources checked
    • cites [129]

      Not checked in the free run.

    • cites [130]

      Not checked in the free run.

    • cites [131] source

      Not checked in the free run.

  • May combine two claims.

    Interpretationcitation

    Algorithmic amplification of emotionally and morally charged content may feed back into human social learning-for instance, inflating beliefs about the normative value of expressing moral outrage, increasing outgroup animosity or creating echo chambers and filter bubbles.

    “The algorithmic amplification of such content may then feed back into human social learning-for instance, inflating beliefs about the normative value of expressing moral outrage 138,139 , increasing outgroup animosity 140 or creating echo chambers and filter bubbles [141][142][143] .”
    Quote in the paperSubstantial claim0 of 3 sources checked
    • cites [138] source

      Not checked in the free run.

    • cites [139] source

      Not checked in the free run.

    • cites [140] source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S4 (based).

    Premisenone given

    Any information-processing system must address a general trade-off between a truthful representation of the raw information and constraints on computation.

    “any information-processing system must address a general trade-off between a truthful representation of the raw information and constraints on computation.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Interpretationreasoning

    Machines' increased computational capacity might affect the feasibility of accumulating uncompressed information via a 'big data' approach.

    “Machines' increased computational capacity might additionally affect the feasibility of accumulating uncompressed information via a 'big data' approach.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S2 (checkable).

    Premisecitation

    These feedback loops, but also deliberate product design choices along with policy approaches, provide potent leverage points for aligning recommender systems with human values.

    “These feedback loops, but also deliberate product design choices along with policy approaches, provide potent leverage points for aligning recommender systems with human values 146 .”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [146] source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S2 (checkable), S3 (load-bearing), S4 (based).

    Interpretationreasoning

    As generative AI becomes more commonplace, distinguishing intrinsic human culture from machine-generated culture or machine-influenced culture becomes even more challenging, especially as watermarking techniques may not be universally adopted.

    “As generative AI becomes more commonplace, distinguishing intrinsic human culture from machine-generated culture or machine-influenced culture becomes even more challenging, especially as watermarking techniques may not be universally adopted.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S2 (checkable), S3 (load-bearing), S4 (based).

    Interpretationreasoning

    It seems likely that machines could recover similar strategies to maintain cultural diversity without human involvement.

    “That said, it seems likely that machines could recover similar strategies to maintain cultural diversity without human involvement.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S2 (checkable), S4 (based). May combine two claims.

    Interpretationnone given

    Since the inception of human culture, derived tools have had an important role in shaping cultural processes, making the establishment of a baseline a challenging question in itself.

    “Since the inception of human culture, derived tools have had an important role in shaping cultural processes, making the establishment of a baseline a challenging question in itself.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Premisecitation

    The language model LLaMA recently gained attention for its relatively smaller size, making it more cost-effective to use.

    “As such, the language model LLaMA recently gained attention for its relatively smaller size, making it more cost-effective to use 167 .”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [167] source

      Not checked in the free run.

  • May combine two claims.

    Interpretationnone given

    Humans select machine behaviour also through direct creation and curation of training data for machine learning, and through more indirect interactions with machine-generated outputs.

    “Humans select machine behaviour also through direct creation and curation of training data for machine learning, and through more indirect interactions with machine-generated outputs.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S2 (checkable), S3 (load-bearing).

    Interpretationreasoning

    Selection is also bound to occur at a level where machines or humans are favoured over one another.

    “Selection is also bound to occur at a level where machines or humans are favoured over one another.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Premisecitation

    Even though present-day machines can conceive messages perceived as more empathic than messages conceived by humans, such messages may be perceived differently once recipients become aware of their artificial origin.

    “For instance, even though present-day machines can conceive messages perceived as more empathic than messages conceived by humans 170,171 , such messages may be perceived differently once recipients become aware of their artificial origin.”
    Quote in the paperNot a substantial claim0 of 2 sources checked
    • cites [170] source

      Not checked in the free run.

    • cites [171] source

      Not checked in the free run.

  • May combine two claims.

    Premisecitation

    This phenomenon stems from a classification bias that favours more prevalent classes, a bias that is amplified through successive iterations of learning.

    “This phenomenon stems from a classification bias that favours more prevalent classes 164 , a bias that is amplified through successive iterations of learning 165 .”
    Quote in the paperSubstantial claim0 of 2 sources checked
    • cites [164] source

      Not checked in the free run.

    • cites [165] source

      Not checked in the free run.

  • May combine two claims.

    Interpretationreasoning

    It therefore seems inevitable that future machines will be trained on mixed human-machine content, forming part of a larger feedback cycle between content generation and selection.

    “It therefore seems inevitable that future machines will be trained on mixed human-machine content, forming part of a larger feedback cycle between content generation and selection.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S2 (checkable), S3 (load-bearing).

    Premisecitation

    As LLMs are trained from curated datasets and fine-tuned using human feedback, quantifying the biases that are introduced and/or mitigated by these models is crucial.

    “However, as LLMs are trained from curated datasets and fine-tuned using human feedback, quantifying the biases that are introduced and/or mitigated by these models is crucial 179,180 .”
    Quote in the paperNot a substantial claim0 of 2 sources checked
    • cites [179] source

      Not checked in the free run.

    • cites [180] source

      Not checked in the free run.

  • May combine two claims.

    Premisecitation

    Despite popular media claims about machine-generated art developing its own unique style, we do not yet have a reliable way of verifying these claims, let alone assessing machine-to-human cultural transmission.

    “Despite popular media claims about machine-generated art developing its own unique style 177 , we do not yet have a reliable way of verifying these claims, let alone assessing machine-to-human cultural transmission 107 .”
    Quote in the paperNot a substantial claim0 of 2 sources checked
    • cites [177] source

      Not checked in the free run.

    • cites [107] source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S4 (based).

    Interpretationnone given

    We currently observe a rapid increase in the diversity of AI models, including LLMs, accelerated by the open-source software movement.

    “We currently observe a rapid increase in the diversity of AI models, including LLMs, accelerated by the open-source software movement.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S2 (checkable).

    Interpretationnone given

    It is difficult to completely disentangle the effects of ranking and recommendation algorithms on culture from alternative processes of human social learning, such as communication technology, institutions and social practices.

    “For instance, it is difficult to completely disentangle the effects of ranking and recommendation algorithms on culture from alternative processes of human social learning, such as communication technology, institutions and social practices.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

    Interpretationnone given

    The ability of machines to process vast amounts of information both quickly and accurately affords them a competitive edge in numerous cognitive tasks, such as strategic gameplay and information retrieval.

    “For instance, the ability of machines to process vast amounts of information both quickly and accurately affords them a competitive edge in numerous cognitive tasks, such as strategic gameplay and information retrieval.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Premisecitation

    Research on 'filter bubbles' has moved from considering algorithmic curation as decisive force shaping online engagement towards acknowledging the influence of users' own choices on social media ranking algorithms and search engines, thus highlighting the intricate feedback loops between machine and human decisions.

    “This challenge is reflected in research on”
    Quote in the paperNot a substantial claim0 of 3 sources checked
    • cites [142] source

      Not checked in the free run.

    • cites [175] source

      Not checked in the free run.

    • cites [176] source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S2 (checkable), S3 (load-bearing), S4 (based). May combine two claims.

    Interpretationnone given

    Selection between humans and machines does not imply that one agent dominates over the other in any cultural niche: often, one will augment rather than fully replace the other; at other times, the presence of the other may trigger the development of new skills or roles.

    “Selection between humans and machines does not imply that one agent dominates over the other in any cultural niche: often, one will augment rather than fully replace the other; at other times, the presence of the other may trigger the development of new skills or roles.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

    Interpretationreasoning

    Human demands towards machines are likely to change over time, shaping the trajectory of machine culture similar to other historic cases where technologies evolved in response to changing human demands.

    “Human demands towards machines are likely to change over time, shaping the trajectory of machine culture similar to other historic cases where technologies evolved in response to changing human demands (for example, the wheel from wooden wheels on carriages to rubber tires on automobiles).”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S4 (based).

    Interpretationnone given

    Market forces, such as regulation and market power, may result in a world dominated by a small number of monolithic models.

    “However, market forces, such as regulation and market power, may result in a world dominated by a small number of monolithic models.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    Chat-GPT outperformed many comparable models due to its superior accessibility and helpfulness.

    “Conversely, Chat-GPT outperformed many comparable models due to its superior accessibility and helpfulness 155 .”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [155]

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Interpretationreasoning

    Across various contexts, humans might favour other humans over machines due to their shared experience.

    “Meanwhile, across various contexts, humans might favour other humans over machines due to their shared experience.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

    Interpretationnone given

    Humans may choose between different intelligent machines available on the market on the basis of factors such as preferences, cost, usefulness, harmfulness and alignment with regulatory requirements.

    “For instance, humans may choose between different intelligent machines available on the market on the basis of factors such as preferences, cost, usefulness, harmfulness and alignment with regulatory requirements.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S2 (checkable), S4 (based). May combine two claims.

    Interpretationreasoning

    Such fragmentation can stifle meaningful dialogue, breed misunderstanding and, ultimately, fracture our shared future vision.

    “Such fragmentation can stifle meaningful dialogue, breed misunderstanding and, ultimately, fracture our shared future vision.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S2 (checkable), S4 (based).

    Interpretationreasoning

    This multi-level selection can, in principle, operate at the level of human organizations augmented by intelligent machines and (eventually) superhuman artificial general intelligence.

    “This multi-level selection can, in principle, operate at the level of human organizations augmented by intelligent machines and (eventually) superhuman artificial general intelligence.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S2 (checkable), S3 (load-bearing), S4 (based).

    Interpretationnone given

    Cultural evolution provides a useful framework for navigating this challenge.

    “Cultural evolution provides a useful framework for navigating this challenge.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S2 (checkable), S4 (based).

    Interpretationnone given

    In all likelihood, the future of culture will be hybrid, with cultural artefacts-scientific theories, industrial processes, art and literature-being created by a combination of human and machine intelligence.

    “In all likelihood, the future of culture will be hybrid, with cultural artefacts-scientific theories, industrial processes, art and literature-being created by a combination of human and machine intelligence.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Premisecitation

    Although experts disagree on the timescales and the degree of risk involved, the potential of superhuman artificial general intelligence poses a possible existential threat to the human species.

    “Although experts disagree on the timescales and the degree of risk involved, the potential of superhuman artificial general intelligence poses a possible existential threat to the human species 192 .”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [192] source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    Another, increasingly used approach is fine-tuning based on RLHF.

    “Another, increasingly used approach is fine-tuning based on RLHF 189 .”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [189]

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S2 (checkable).

    Premisecitation

    Ensuring that AI models, such as LLMs, reflect the beliefs and values of a given community requires mechanisms for societal decision-making about what knowledge goes into the models.

    “Ensuring that AI models, such as LLMs, reflect the beliefs and values of a given community requires mechanisms for societal decision-making about what knowledge goes into the models 181 .”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [181] source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S2 (checkable), S3 (load-bearing).

    Interpretationcitation

    These dynamics may lead to diverse phenomena ranging from different forms of human-machine mutualism to Red Queen effects (an evolutionary arms race between humans and machines) that characterize the co-evolution of both forms of intelligence.

    “These dynamics may lead to diverse phenomena ranging from different forms of human-machine mutualism to Red Queen effects (an evolutionary arms race between humans and machines) that characterize the co-evolution of both forms of intelligence 191 .”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [191] source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S4 (based).

    Interpretationreasoning

    This raises the possibility of reduction in cultural diversity, as major social, political and economic forces try to shape global machine culture to match their preferences.

    “This raises the possibility of reduction in cultural diversity, as major social, political and economic forces try to shape global machine culture to match their preferences.”
    Quote in the paperNot a substantial claim
    • Rests on an assumption: Cultural diversity is valuable and should be preserved.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    Yet another approach is to fine-tune using constitutional rules provided by humans.

    “Yet another approach is to fine-tune using constitutional rules provided by humans 190 .”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [190] source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S2 (checkable), S3 (load-bearing), S4 (based).

    Interpretationnone given

    The relative merits and drawbacks of these various approaches are still not well understood.

    “The relative merits and drawbacks of these various approaches are still not well understood.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    One approach to ensure cultural beliefs and values are present in an AI model is to carefully curate the dataset on which the model is trained.

    “One approach is to carefully curate the dataset on which the model is trained 187,188 .”
    Quote in the paperNot a substantial claim0 of 2 sources checked
    • cites [187] source

      Not checked in the free run.

    • cites [188] source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S1 (asserted), S2 (checkable), S3 (load-bearing).

    Premisecitation

    Similar to the notion of human-in-the-loop control of intelligent machines, we can aspire towards society-in-the-loop control of the complex phenomena of machine culture.

    “Similar to the notion of human-in-the-loop control of intelligent machines, we can aspire towards society-in-the-loop control of the complex phenomena of machine culture 186 .”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [186] source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Premisecitation

    A number of interesting ideas are emerging, from voting on different algorithmic policymakers to using LLMs to summarize diverse human opinions and generate consensus statements.

    “A number of interesting ideas are emerging, from voting on different algorithmic policymakers 184 to using LLMs to summarize diverse human opinions 185 and generate consensus statements 153 .”
    Quote in the paperNot a substantial claim0 of 3 sources checked
    • cites [184]

      Not checked in the free run.

    • cites [185] source

      Not checked in the free run.

    • cites [153] source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S2 (checkable), S4 (based).

    Interpretationreasoning

    As AI models become increasingly personalized, conforming to and reinforcing our individual worldviews, they risk engendering an unprecedented fragmentation of our shared perception of the world.

    “As AI models become increasingly personalized, conforming to and reinforcing our individual worldviews, they risk engendering an unprecedented fragmentation of our shared perception of the world.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S2 (checkable).

    Interpretationreasoning

    There is a need for methods to check which cultural beliefs and values have been learned, inferred or encoded in a given AI model and the degree to which they align with a target culture.

    “There is a need for methods to check which cultural beliefs and values have been learned, inferred or encoded in a given AI model and the degree to which they align with a target culture.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S2 (checkable), S4 (based).

    Interpretationreasoning

    If we continually interact with machines that echo and affirm our preconceived notions, we risk isolating ourselves within ideologically and culturally homogenous echo chambers.

    “Drawing a parallel, if we continually interact with machines that echo and affirm our preconceived notions, we risk isolating ourselves within ideologically and culturally homogenous echo chambers.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S4 (based).

    Interpretationnone given

    This process may be amplified by feedback loops, in which LLMs train on an ongoing basis from synthetic data or from human data that contains much machine-generated text.

    “This process may be amplified by feedback loops, in which LLMs train on an ongoing basis from synthetic data or from human data that contains much machine-generated text.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S2 (checkable), S4 (based).

    Interpretationreasoning

    Engineering this evolutionary process can provide means for ensuring human survival and agency in the long run.

    “Engineering this evolutionary process can provide means for ensuring human survival and agency in the long run.”
    Quote in the paperNot a substantial claim
    • Rests on an assumption: Human survival and agency are desirable goals.
87 of 87

  • The paper relies on this without support.

    The three Darwinian properties (variation, transmission, selection) are sufficient to capture the ways machines influence cultural evolution.

    “These properties are not necessarily the result of distinct processes of human cognition and behaviour, yet they offer a useful framework to analyse the multifaceted ways machines can influence cultural evolution (see Table 1 for a summary).”

  • The paper relies on this without support.

    The examples of machine impact (AlphaGo, LLMs, recommender systems) are representative of broader machine culture.

    “Building on these exemplary instances of how machine technologies may impact cultural evolution, we will now map out a systematic framework for studying the potential of machines to shape cultural evolutionary processes.”

  • The paper relies on this without support.

    Current trends in AI development will continue.

    “As the capabilities and uses of LLMs continue to develop, the value of certain human skills will shift.”

  • The paper relies on this without support.

    Cultural diversity is valuable and should be preserved.

    “This raises the possibility of reduction in cultural diversity, as major social, political and economic forces try to shape global machine culture to match their preferences.”
    “It is also imperative to understand what constitutes a 'healthy' level of diversity, one that retains local sovereignty while also fostering collective human flourishing.”

  • The paper relies on this without support.

    Human survival and agency are desirable goals.

    “Engineering this evolutionary process can provide means for ensuring human survival and agency in the long run.”
5 of 5

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?

  • Our analysis can inform the debate around LLM policies by providing a measurement method for LLM usage.

    “Our analysis can inform the necessary debate around LLM policies by providing a measurement method for LLM usage that is urgently needed (52, 53).”

    From Delving into LLM-assisted writing in biomedical publications through excess vocabulary (2025) · cites this paper as [52]

    The Measurement section (page unavailable) calls for better metrics of machines’ cultural impact and describes quantification as an open challenge. That supports the need for measurement, but the paper does not specifically call for a method to measure LLM usage; its focus is cultural effects, not usage rates. The citing authors’ proposed method is their own contribution, not a claim this source makes. (LLM, conf=0.82)

    Experimental — accuracy not yet measured

  • Researchers highlight the homogenization risks posed by overreliance on AI, which can reinforce biases and reduce narrative diversity.

    “researchers highlight the homogenization risks posed by overreliance on AI (Bender et al., 2021;Brinkmann et al., 2023;Doshi & Hauser, 2024), which can reinforce biases and reduce narrative diversity.”

    From Fanfiction in the Age of AI: Community Perspectives on Creativity, Authenticity and Adoption (2026) · cites this paper as Brinkmann et al., 2023

  • Generative AI is more appropriately conceptualized as a cultural system embedded with human assumptions, institutional agendas, and structural asymmetries than as a neutral technological tool.

    “Rather than viewing generative AI as a neutral technological tool, it is more appropriate to conceptualize it as a cultural system embedded with human assumptions, institutional agendas, and structural asymmetries (Brinkmann et al., 2023;Nick, 2017).”

    From Fanfiction in the Age of AI: Community Perspectives on Creativity, Authenticity and Adoption (2026) · cites this paper as Brinkmann et al., 2023

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

Each contradiction between two passages listed here was confirmed by an independent check.

  • Collaborative filtering points to the wrong figure

    Minor

    The text cites Fig. 3b for collaborative filtering, but Fig. 3b shows Go players' decision quality. The collaborative-filtering illustration is labelled Fig. 4b.

1 of 1

Cites

Flagged first, then by citations

Research works printed but not linked: 22

  1. Mokyr, J. A Culture of Growth: The Origins of the Modern Economy (Princeton Univ. Press, 2017). · DOI
  2. Needham, J. in Chemistry and Chemical Technology, Pt. 7: Military Technology—the Gunpowder Epic Vol. 5 (Cambridge Univ. Press, 1986).
  3. Russell, S. & Norvig, P. Artificial Intelligence: A Modern Approach (Prentice Hall, 2009).
  4. Kurzweil, R., Richter, R., Kurzweil, R. & Schneider, M. L. The Age of Intelligent Machines (MIT Press, 1990).
  5. Ramesh, A. et al. Zero-shot text-to-image generation. In International Conf. on Machine Learning 8821–8831 (PMLR, 2021).
  6. Mikolov, T., Yih, W. & Zweig, G. Linguistic regularities in continuous space word representations. In Proc. 2013 Conf. of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies 746–751 (Association for Computational Linguistics, 2013).
  7. Wang, Y., Shimada, K. & Barati Farimani, A. Airfoil GAN: encoding and synthesizing airfoils for aerodynamic shape optimization. J. Comput. Des. Eng. 10, 1350–1362 (2023).
  8. Shin, M., Kim, J. & Kim, M. Human learning from artificial intelligence: evidence from human Go players’ decisions after AlphaGo. Proc. Annu. Meet. Cogn. Sci. Soc. 43, 43 (2021).
  9. Boyd, R. & Richerson, P. J. Culture and the Evolutionary Process (Univ. of Chicago Press, 1985).
  10. Aveni, A. F. Skywatchers: A Revised and Updated Version of Skywatchers of Ancient Mexico (Univ. of Texas Press, 2001).
  11. Brown, T. et al. Language models are few-shot learners. Adv. Neural Inf. Process. Syst. 33, 1877–1901 (2020).
  12. Buolamwini, J. & Gebru, T. Gender shades: intersectional accuracy disparities in commercial gender classification. In Proc. 1st Conference on Fairness, Accountability and Transparency 77–91 (PMLR, 2018).
  13. Bartlett, F. C. Remembering: A Study in Experimental and Social Psychology xix, 317 (Cambridge Univ. Press, 1932).
  14. Sharma, D. K. & Sharma, A. A comparative analysis of web page ranking algorithms. Int. J. Comput. Sci. Eng. 2, 2670–2676 (2010).
  15. Tkalcic, M., Kosir, A. & Tasic, J. Affective recommender systems: the role of emotions in recommender systems. In The RecSys 2011 Workshops-Decisions@ RecSys 2011 and UCERSTI-2: Human Decision Making in Recommender Systems
  16. User-Centric Evaluation of Recommender Systems and Their Interfaces-2 Vol. 811, 9-13 (CEUR-WS.org, 2011).
  17. Zheng, Y., Mobasher, B. & Burke, R. D. The role of emotions in context-aware recommendation. Decis. RecSys 2013, 21–28 (2013).
  18. Brady, W. J., Jackson, J. C., Lindström, B. & Crockett, M. J. Algorithm-mediated social learning in online social networks. Trends Cogn. Sci. (in the press).
  19. Sunstein, C. R. Republic.com 2.0 (Princeton Univ. Press, 2007).
  20. Brandt, F., Conitzer, V. & Endriss, U. in Multiagent Systems (ed. Weiss, G.) 213–284 (MIT Press, 2012).
  21. Laurençon, H. et al. The bigscience roots corpus: a 1.6 tb composite multilingual dataset. Adv. Neural Inf. Process. Syst. 35, 31809–31826 (2022).
  22. Bostrom, N. Superintelligence: Paths, Dangers, Strategies (Oxford Univ. Press, 2014).

Cited by

Sorted by citations

Related

Authors
Brinkmann, L., Baumann, F., Bonnefon, J.-F., Derex, M., Müller, T. F., Nussberger, A.-M., Czaplicka, A., Acerbi, A., Griffiths, T. L., Henrich, J., Leibo, J. Z., McElreath, R., Oudeyer, P.-Y., Stray, J., & Rahwan, I.
Affiliations
  • Max Planck Institute for Human Development (DE)
  • Toulouse School of Economics (FR)
  • University of Trento (IT)
  • Princeton University (US)
  • Harvard University (US)
  • Google DeepMind (United Kingdom) (GB)
  • Max Planck Institute for Evolutionary Anthropology (DE)
  • Institut national de recherche en sciences et technologies du numérique (FR)
  • University of California, Berkeley (US)
Corresponding author
  • Brinkmann, Levin
  • Rahwan, Iyad
ORCID
Venue
Nature Human Behaviour
Publisher
Springer Science and Business Media LLC
Published
2023-11-20
Volume
7
Issue
11
Pages
1855-1868
ISSN
2397-3374 (ISSN-L)
DOI
10.1038/s41562-023-01742-2
OpenAlex
W4388835050
Semantic Scholar
eb1951e9d070b73817e328e589220893c2b108a2
Cited by
132
References
171
Citations by year
2026: 54 · 2025: 51 · 2024: 23 · 2023: 1 · 2022: 3
FWCI
16.81
Type
journal article
Language
en
Open access
Licence
Locations
  • Nature Human Behaviour (publishedVersion)
  • PubMed (publishedVersion)
  • Toulouse 1 Capitole Publications (Université Toulouse I Capitole) (acceptedVersion)
  • arXiv (Cornell University) (submittedVersion, cc-by-nc-nd)
  • HAL (submittedVersion, other-oa)
  • Iris (University of Trento) (submittedVersion, other-oa)
  • arXiv (Cornell University)
Funders
Agence Nationale de la Recherche (ANR-17-EURE-0010, ANR-19-P3IA-0004, ANR-17-EURE-, 17-EURE-00-10, ANR-17)
Topics
  • Social Sciences › Cultural Studies › Language and cultural evolution
  • Social Sciences › Arts and Humanities › History and Philosophy of Science › Cybernetics and Technology in Society
  • Social Sciences › Safety Research › Ethics and Social Impacts of AI
Keywords
machine culture, cultural evolution, intelligent machines, recommendation algorithms, social learning, chatbots, cultural transmission, cultural models, computer art
MeSH terms
Animals, Cultural Evolution, Culture, Humans, Learning, Hominidae

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.

Claims read by deepseek-v4.1-flash under definition version 3 (extract-v3) on 2026-10-10.

Classified by jev-1.13-20260917 (gold-guideline-v0.2) on 2026-10-10.

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