Thirty years of work in cultural evolution, primatology, human behavioral ecology, and cognitive science has established a consensus framework for understanding culture. This framework characterizes agents as voracious epistemic optimizers: Individuals who exploit cues, adopt strategies, and intervene on situations to extract high-quality information relative to their goals. This framing extends out into the world, seeing it in epistemic terms: the physical environment and other agents are repositories of, and instruments for, information acquisition.
Veissière et al. (hereafter, “the authors”) adopt and synthesize this consensus framework. Their particular concern is the thoroughly social character of the informational world in which humans develop and live – and their novel contribution is to wed empirical research on this topic with the apparatuses of variational Bayesian inference and the free-energy principle. These tools, they suggest, provide means of modeling key features of enculturation, behavior, and cultural change.
Yet, an important feature of this account needs to be noted at the outset. The authors’ epistemic framing grounds both culture and enculturation in the extraction and employment of information, and in so doing, minimizes the explanatory clout of other core aspects of human life; notably, deliberative choice, affect, and normativity. On the authors’ account, these latter features are either reducible or subsidiary to variational inference. The result is a conservative model of culture, one already well-formulated by David Laitin (Reference Laitin2007): cultures are “circumstances in which members of a group […] are able to condition their behaviour on common knowledge beliefs about the behaviour of all members of the group” (p. 64). Such a model renders cultures largely homeostatic; enculturation means both learning and expecting others to stay within the bounds of established behavior.
No doubt models couched within this framework are useful for understanding cultural phenomena; for instance, explaining how adolescents flexibly adopt different learning strategies to become capable members of the group (e.g., Salali et al. Reference Salali, Chaudhary, Bouer, Thompson, Vinicius and Migliano2019). But, as a general account of culture, the picture is likely impoverished. Enculturated agents do not just think through other agents in an epistemic way, as something to be conditioned over, but in a normative and affective way: A confession of love can be terrifying even before hearing the response; and even thinking about committing a crime can bring feelings of disquiet and shame. Norms, in other words, are not mere conventions – subject to more or less accurate conditionalization – but, affectively motivated standards for appropriate behavior. More importantly, they are ubiquitous structures of human sociality.
Normative and affective features are difficult to grasp using the language of epistemic optimization. Consider a key piece of the authors’ machinery; cultural, or “epistemic” affordances. Affordances are relationships between features in the environment and agential behavioral repertoire (Chemero Reference Chemero2009; Walsh Reference Walsh2015). Although some researchers do link affordances to normatively freighted concepts such as “skill” (e.g., Rietveld & Kiverstein Reference Rietveld and Kiverstein2014), this is not because the two are identical. They are not. Indeed, the ubiquity of human norm-governedness is a puzzle, and there is a growing literature aimed at exploring just how affordance-structured ape cognition could be bootstrapped and changed into norm-rich human cognition (Birch, Reference Birchm.s.).
Though the authors do mention normativity, it is as a fait accompli – something that comes for free with variational inference in a structured environment. But, this is where one is owed further explanation: How are affordances bootstrapped into skills? How does variational inference explain how human beings are motivated to act in accordance with rules? In short, how epistemic optimizing explain norm-governed behavior? Although the authors show some sensitivity to these issues, they owe us an account that explains the development of such capacities in inferential terms.
This line of thinking leads to broader concerns with how the Bayesian framework construes and understands culture. Cultural groups are governed by local normative structures that set standards for behavior and regulate social interaction. Yet these normative frameworks are continually contested and subject to change, and this change often involves deliberate attempts on the part of individuals to reorganize governing structures. Whether this involves shifts in kinship or marriage practices, corporate or political governance, religious or ritual practices, agents often act collectively to change the normative strictures guiding their behavior. Yet the epistemic account on offer leaves little explanatory space to account for such social change. This is especially true of the core organizational structures of human sociality.
The authors’ account of social change ties it to exploration and the acquisition of information. Agents are motivated by “intrinsic, epistemic imperatives” to acquire information before engaging in actions with pragmatic value – turning on the lights to avoid stumbling on the furniture, for instance. Nonetheless, the authors suggest, these epistemic imperatives have the added benefit of promoting curiosity and by extension, enlarging the “repertoire of responses for the individual or the group” (sect. 3.3, para. 6).
Yet, even if human beings were motivated by such imperatives, this could not be the whole story. For many core organizational features of human life are exclusive: it is either possible to marry your cross-cousin or impossible; you either pay brideprice or a dowry; you either live in a big man society or you do not. Transitions between these organizational states are complex – and are often the result of deliberative, communal experimentation (Wengrow & Graeber Reference Wengrow and Graeber2015) – but it is difficult to see how epistemic imperatives and enlarged “repertoires of responses” get a grip on these changes. What would it mean to get information about breaking taboos on marrying cross-cousins without actually marrying a cross-cousin? What about motivating systematic change in governance? How could one motivate pragmatic action without knowing what comes next?
As noted, the authors’ account is conservative – focusing on how to things keep going on as before. But, radical organizational change seems to be ubiquitous among human cultures (Mace & Jordan Reference Mace and Jordan2011), suggesting that the authors’ model should be open to such phenomena. Yet, such openness will require greater attention to the role of affect, normativity, and communal decision-making – and this in turn may require going beyond Bayesian variational inference and the epistemic framing of human beings and the world.
Samuel P. L. Veissière
[Opens in a new window]Samuel P. L. Veissière
,
Thirty years of work in cultural evolution, primatology, human behavioral ecology, and cognitive science has established a consensus framework for understanding culture. This framework characterizes agents as voracious epistemic optimizers: Individuals who exploit cues, adopt strategies, and intervene on situations to extract high-quality information relative to their goals. This framing extends out into the world, seeing it in epistemic terms: the physical environment and other agents are repositories of, and instruments for, information acquisition.
Veissière et al. (hereafter, “the authors”) adopt and synthesize this consensus framework. Their particular concern is the thoroughly social character of the informational world in which humans develop and live – and their novel contribution is to wed empirical research on this topic with the apparatuses of variational Bayesian inference and the free-energy principle. These tools, they suggest, provide means of modeling key features of enculturation, behavior, and cultural change.
Yet, an important feature of this account needs to be noted at the outset. The authors’ epistemic framing grounds both culture and enculturation in the extraction and employment of information, and in so doing, minimizes the explanatory clout of other core aspects of human life; notably, deliberative choice, affect, and normativity. On the authors’ account, these latter features are either reducible or subsidiary to variational inference. The result is a conservative model of culture, one already well-formulated by David Laitin (Reference Laitin2007): cultures are “circumstances in which members of a group […] are able to condition their behaviour on common knowledge beliefs about the behaviour of all members of the group” (p. 64). Such a model renders cultures largely homeostatic; enculturation means both learning and expecting others to stay within the bounds of established behavior.
No doubt models couched within this framework are useful for understanding cultural phenomena; for instance, explaining how adolescents flexibly adopt different learning strategies to become capable members of the group (e.g., Salali et al. Reference Salali, Chaudhary, Bouer, Thompson, Vinicius and Migliano2019). But, as a general account of culture, the picture is likely impoverished. Enculturated agents do not just think through other agents in an epistemic way, as something to be conditioned over, but in a normative and affective way: A confession of love can be terrifying even before hearing the response; and even thinking about committing a crime can bring feelings of disquiet and shame. Norms, in other words, are not mere conventions – subject to more or less accurate conditionalization – but, affectively motivated standards for appropriate behavior. More importantly, they are ubiquitous structures of human sociality.
Normative and affective features are difficult to grasp using the language of epistemic optimization. Consider a key piece of the authors’ machinery; cultural, or “epistemic” affordances. Affordances are relationships between features in the environment and agential behavioral repertoire (Chemero Reference Chemero2009; Walsh Reference Walsh2015). Although some researchers do link affordances to normatively freighted concepts such as “skill” (e.g., Rietveld & Kiverstein Reference Rietveld and Kiverstein2014), this is not because the two are identical. They are not. Indeed, the ubiquity of human norm-governedness is a puzzle, and there is a growing literature aimed at exploring just how affordance-structured ape cognition could be bootstrapped and changed into norm-rich human cognition (Birch, Reference Birchm.s.).
Though the authors do mention normativity, it is as a fait accompli – something that comes for free with variational inference in a structured environment. But, this is where one is owed further explanation: How are affordances bootstrapped into skills? How does variational inference explain how human beings are motivated to act in accordance with rules? In short, how epistemic optimizing explain norm-governed behavior? Although the authors show some sensitivity to these issues, they owe us an account that explains the development of such capacities in inferential terms.
This line of thinking leads to broader concerns with how the Bayesian framework construes and understands culture. Cultural groups are governed by local normative structures that set standards for behavior and regulate social interaction. Yet these normative frameworks are continually contested and subject to change, and this change often involves deliberate attempts on the part of individuals to reorganize governing structures. Whether this involves shifts in kinship or marriage practices, corporate or political governance, religious or ritual practices, agents often act collectively to change the normative strictures guiding their behavior. Yet the epistemic account on offer leaves little explanatory space to account for such social change. This is especially true of the core organizational structures of human sociality.
The authors’ account of social change ties it to exploration and the acquisition of information. Agents are motivated by “intrinsic, epistemic imperatives” to acquire information before engaging in actions with pragmatic value – turning on the lights to avoid stumbling on the furniture, for instance. Nonetheless, the authors suggest, these epistemic imperatives have the added benefit of promoting curiosity and by extension, enlarging the “repertoire of responses for the individual or the group” (sect. 3.3, para. 6).
Yet, even if human beings were motivated by such imperatives, this could not be the whole story. For many core organizational features of human life are exclusive: it is either possible to marry your cross-cousin or impossible; you either pay brideprice or a dowry; you either live in a big man society or you do not. Transitions between these organizational states are complex – and are often the result of deliberative, communal experimentation (Wengrow & Graeber Reference Wengrow and Graeber2015) – but it is difficult to see how epistemic imperatives and enlarged “repertoires of responses” get a grip on these changes. What would it mean to get information about breaking taboos on marrying cross-cousins without actually marrying a cross-cousin? What about motivating systematic change in governance? How could one motivate pragmatic action without knowing what comes next?
As noted, the authors’ account is conservative – focusing on how to things keep going on as before. But, radical organizational change seems to be ubiquitous among human cultures (Mace & Jordan Reference Mace and Jordan2011), suggesting that the authors’ model should be open to such phenomena. Yet, such openness will require greater attention to the role of affect, normativity, and communal decision-making – and this in turn may require going beyond Bayesian variational inference and the epistemic framing of human beings and the world.