Heyes’ (Reference Heyes2018) Cognitive Gadgets presents a compelling and erudite case for the influence of cultural evolution on the emergence of the distinctive cognitive gadgets unique to humans. Her argument is a convincing one – she offers a lucid treatment of four empirically supported examples to substantiate her claims and highlights a number of viable research avenues to open her theory to scientific scrutiny. Cognitive Gadgets no doubt stands to make a meaningful contribution to current thinking about the evolutionary roots of human cognition. Admittedly, our opinion is also biased – cultural evolutionary psychology resonates with our own work, and we hope to see its widespread adoption as a research program.

There are, however, a number of points worth raising about Heyes’ misleading characterisation of evolutionary psychology. First, the decidedly gene-centric view she attributes to members of the “High Church” of evolutionary psychology is somewhat specious – to understand and model gene-environment interactions and the ways in which social and physical ecologies impact the development of human cognitive specialisations is a core commitment of the field (Geary & Bjorklund Reference Geary and Bjorklund2000; Krebs Reference Krebs2003; Ploeger et al. Reference Ploeger, van der Maas and Raijmakers2008). Heyes also overlooks the heterogeneity of evolutionary psychology, broadly construed. For example, consistent with her own teleosemantic perspective, proponents of an “evolutionary systems” approach adopt an expanded view of inheritance, which extends beyond the gene to incorporate other forms of intergenerational information transmission, including epigenetic effects and the inheritance of cultural artefacts and patterned cultural practices (Badcock Reference Badcock2012; Caporael Reference Caporael2001; Hendriks-Jansen Reference Hendriks-Jansen1996; Laland Reference Laland2017; Lickliter & Honeycutt Reference Lickliter and Honeycutt2003; Ramstead et al. Reference Ramstead, Veissière and Kirmayer2016; Roepstorff et al. Reference Roepstorff, Niewöhner and Beck2010).

By emphasising Chomsky's (Reference Chomsky1965) “poverty of stimulus” criterion and our innate cognitive “starter kit,” Heyes also creates the impression that evolved, biological traits that help us adapt to the social milieu should largely be relegated to early biases, which are then refined by nurture and culture. However, there are clearly many adaptive traits that show a strong biological basis and profoundly affect social cognition and behaviour throughout the lifespan. Obvious examples include the gross morphology and physiological properties of the brain (Friston Reference Friston2010); hormonal and neuromodulatory systems (Katz & Harris-Warrick Reference Katz and Harris-Warrick1999; McGlothlin & Ketterson Reference McGlothlin and Ketterson2008); reward, mood, and affective systems (Gray Reference Gray, Ekman and Davidson1994; Nettle & Bateson Reference Nettle and Bateson2012); personality traits (Bouchard & Loehlin Reference Bouchard and Loehlin2001; Nettle Reference Nettle2006); and sensitive periods of development (e.g., puberty) that fine-tune our adaptation to different socio-environmental contexts across the life course (Frankenhuis & Fraley Reference Frankenhuis and Fraley2017; Geary & Bjorklund Reference Geary and Bjorklund2000).

More critically, Cognitive Gadgets tends to neglect the fundamental role of biology in shaping our cultural worlds. Dynamical simulation studies have provided proof of principle that individual differences in adaptive decision rules (e.g., mating or social learning strategies) create marked changes in the self-organisation of social norms and cultural dynamics (Kenrick et al. Reference Kenrick, Li and Butner2003; Molleman et al. Reference Molleman, Van den Berg and Weissing2014). More substantively, evolutionary psychologists have accumulated a wealth of evidence to suggest that cognitive traits favoured by natural selection exert a powerful influence on the sorts of cultural expectations, norms, and practices that are likely to evolve – including those surrounding communal sharing and morality (Kameda et al. Reference Kameda, Takezawa and Hastie2005; Krebs Reference Krebs2008), in-group versus out-group behaviours (Brewer Reference Brewer2007), and behaviours involving social exchange (Wischniewski et al. Reference Wischniewski, Windmann, Juckel and Brüne2009). Such work reminds us that culture shouldn't be individuated from the evolved biobehavioural dynamics of the individuals that comprise it (Kenrick et al. Reference Kenrick, Li and Butner2003; Lehman et al. Reference Lehman, Chiu and Schaller2004). The idea that we gain remarkable cognitive capacities via cultural learning almost goes without saying – but this should not detract from a dialectical view that sees biological and cultural inheritance as mutually constitutive.

As such, our main reservation with Cognitive Gadgets is that it promotes a sharp distinction between nature, nurture, and culture. As evidenced by our own work in this area (Constant et al. Reference Constant, Ramstead, Veissiere, Campbell and Friston2018; Kirmayer & Ramstead Reference Kirmayer, Ramstead, Durt, Fuchs and Tewes2017; Ramstead et al. Reference Ramstead, Veissière and Kirmayer2016; Reference Ramstead, Constant, Badcock and Friston2019), we certainly agree that cultural influences play an essential role in both the inheritance and development of our adaptive cognitive specialisations. However, we do not think that these three sources can be pried apart so easily – our interest lies more in the ways they interact to produce human phenotypes. Human biology is also a cultural biology; and human culture is realised by interacting biological systems within a shared material niche (Kirmayer & Ramstead Reference Kirmayer, Ramstead, Durt, Fuchs and Tewes2017). Indeed, for more than 200,000 years, the main selection pressure on human survival has been the capacity to access and leverage accumulated sociocultural information (Henrich Reference Henrich2015; Hrdy Reference Hrdy2009; Tomasello Reference Tomasello2014). Heyes wouldn't deny this – she proposes that humans begin with a genetically specified “starter kit” that is geared towards navigating the sociocultural world, which allows more sophisticated forms of social cognition, like literacy and mindreading, to develop. Arguably, such “innate” propensities only ever emerged because of the increasing importance of sociocultural information for human survival.

Accordingly, we think that one of the most interesting questions raised by Cognitive Gadgets relates to the mechanisms that explain how biological, social, and cultural dynamics interact. This appeals to an evolutionary systems perspective that is able to capture both the ways in which biological and cultural evolution shape individual minds and the ways in which individual minds shape culture and biology (Badcock Reference Badcock2012; Caporael Reference Caporael2001; Kenrick Reference Kenrick2001). Of course, such multilevel dynamics are challenging to understand, let alone study.

Fortunately, a promising approach has emerged from neuroscience and theoretical biology that has the potential to provide such a multiscale modelling strategy, called active inference. A descendent of predictive coding schemes of the brain (see Clark Reference Clark2013), active inference is a mathematical formulation that describes how living systems are able to maintain themselves within a limited range of phenotypic states, that is, within the set of states in which they expect to find themselves, on average and over time. It explains how biological systems appear to resist the natural tendency to dissipate into their environment by fulfilling biologically instantiated (Bayesian) prior beliefs or expectations about the way the world unfolds (Friston Reference Friston2010; Friston et al. Reference Friston, Daunizeau and Kiebel2009). In short, organisms are driven by the biological imperative to maintain homeostasis via action-perception loops that actively minimise “surprise.” This framework has recently been extended to explain the evolution, development, and multiscale dynamics of living systems in general (Friston Reference Friston2013; Kirchhoff et al. Reference Kirchhoff, Parr, Palacios, Friston and Kiverstein2018; Ramstead et al. Reference Ramstead, Badcock and Friston2018). Here, we concentrate on two of our own complementary approaches derived from active inference, which we believe are particularly relevant to Heyes’ proposal.

On the one hand, Heyes’ appeal to our unique cognitive specialisations connects with a new theory of the human brain, cognition, and behaviour called the hierarchically mechanistic mind (HMM); see Badcock et al. (Reference Badcock, Friston and Ramstead2019a; Reference Badcock, Friston, Ramstead, Ploeger and Hohwy2019b). This model rests on two fundamental claims. The first follows active inference by suggesting that the brain is a complex adaptive system comprising hierarchically organised neurocognitive mechanisms that function to reduce the dispersion or decay of our sensory and physical states by producing action-perception cycles that seek to minimise surprise (Badcock et al. Reference Badcock, Friston and Ramstead2019a). The second claim follows Tinbergen's (Reference Tinbergen1963) four questions in ethology (i.e., adaptation, phylogeny, ontogeny, and mechanism) by suggesting that neural form and function can only be understood in terms of the broader evolutionary, intergenerational, developmental, and real-time processes that act on human phenotypes, which are differentially illuminated by major paradigms in psychology (i.e., evolutionary psychology, evo-devo, developmental psychology, and psychology's subdisciplines, respectively; Badcock et al. Reference Badcock, Friston, Ramstead, Ploeger and Hohwy2019b). Thus, to understand a phenotypic trait, researchers need to develop multiscale hypotheses that synthesise findings from diverse fields of psychological inquiry to explain both why that trait is adaptive and how it emerges from evolutionary, intergenerational, developmental, and real-time processes (for an application to depression, see Badcock et al. Reference Badcock, Davey, Whittle, Allen and Friston2017). The HMM situates Heyes’ work within a broader meta-theory of psychological inquiry that sees cultural evolutionary psychology as but one viable approach to understanding the evolution of human traits – a paradigm, like evo-devo, that concentrates on the group-level, intergenerational dynamics that bridge human evolution and development, thereby driving phylogenetic change. This complements insights from other evolutionary paradigms; it certainly doesn't contradict them.

To borrow Heyes’ own term, active inference also supplies a plausible “force theory” of the nested socio-environmental dynamics responsible for the evolution and development of cognitive gadgets at the level of the individual. This brings us to our second complementary approach, called the variational approach to niche construction (VANC) (Constant et al. Reference Constant, Ramstead, Veissiere, Campbell and Friston2018).

The VANC considers niche construction, that is, implicit and explicit modifications of the environment, as a corollary of active inference, whereby embrained expectations guiding adaptive action-perception loops come to be encoded in the material layout of human niches. Take, for instance, desire paths. As they cut through a grassy field on their daily commute, people implicitly leave traces that inform other pedestrians of the possible intentions of those who crossed the path before; for example, “I want to reach the eastern exit of the park.” By engaging the well-worn path, a novice agent can zero in on the optimal route without having any knowledge of the park's design. Our approach here is consistent with Heyes’ – all this agent would require is the propensity to let herself be guided by the path. This disposition is made possible, presumably, by a minimal starter kit that includes basic cooperative sensory and motor dispositions, as well as some expectations regarding what she herself desires to do; for example, “I expect that my action will lead me to the eastern exit.” By engaging the desire path, the agent will further wear down the trail, thereby increasing its reliability for others. Of course, another example of niche construction, which relates directly to Heyes’ treatment of language, is written text.

The point here is that, through niche construction, humans produce culturally specific behavioural patterns encoded in the constructed artefacts that populate their niche, which they can then recruit to support the performance of various tasks. In so doing, they often implicitly and automatically converge on statistically recurrent behaviours, which, following active inference, are the least surprising ones (i.e., those that characterise the local cultural phenotype) (Constant et al. Reference Constant, Ramstead, Veissiere, Campbell and Friston2019). Over evolutionary time, cultural evolution scaffolds and finesses progressively complex “nurtural” networks of externally realised expectations (in patterned cultural practices and constructed niches), thereby guiding and transmitting increasingly sophisticated cultural behaviour (Constant et al. Reference Constant, Ramstead, Veissiere, Campbell and Friston2018; Ramstead et al. Reference Ramstead, Veissière and Kirmayer2016; Veissière et al. Reference Veissière, Constant, Ramstead, Friston and Kirmayer2019).

Under both the HMM and VANC, cognitive gadgets can be described as heritable adaptive priors that underlie (neuronally encoded) expectations about the dynamics of the social world and guide our action-perception cycles towards unsurprising states (see Badcock et al. Reference Badcock, Friston, Ramstead, Ploeger and Hohwy2019b). According to this perspective, such priors have emerged from the reciprocal interplay of biological, sociocultural, and ecological dynamics over evolutionary time because they have afforded a reliable means to reduce socio-environmental uncertainty. In other words, cognitive gadgets can be thought to entail hierarchical architectures of adaptive prior expectations encoded at multiple levels and sites, spanning neural systems, human phenotypes, social interactions, culturally specified motor patterns, and ecological structures.

Arguably, this notion adds to Heyes’ proposal in two important ways. First, it avoids the questionable claim that our distinctive cognitive faculties are chiefly cultural products by suggesting that nature, nurture, and culture operate synergistically to optimise our phenotypes and eco-niches over evolutionary, intergenerational, and developmental timescales. Shedding light on such faculties requires recourse to research that spans the full breadth of evolutionary psychology (cultural or otherwise), not to mention allied disciplines like anthropology, biology, and ecology. Second, by subsuming active inference, the HMM and VANC afford a single, common language that allows us to describe both biological and cultural influences on human phenotypes mathematically, to model their interactions computationally, to test these models via simulation studies, and then to compare the outcomes of such in silico research with real-world experiments and observations (Badcock et al. Reference Badcock, Friston and Ramstead2019a; Ramstead et al. Reference Ramstead, Badcock and Friston2018).

In sum, our approach to human cognition builds on that of Heyes, but blurs the lines between nature, nurture, and culture – proposing instead a single, generic information theoretic mechanism (i.e., active inference) that expresses itself in different, complementary ways across all three. With this in mind, we suggest that active inference would make a powerful addition to the explanatory toolbox of cultural evolutionary psychology.