Binz et al. highlight the potential of meta-learning to greatly enhance the flexibility of AI algorithms, as well as to approximate human behavior more accurately than traditional learning methods. We wish to emphasize a basic problem that lies underneath these two objectives, and in turn suggest another perspective of the required notion of “meta” in meta-learning: knowing what to learn.

Recent studies suggest that meta-learning may provide an original solution to an enduring puzzle about whether neural networks can explain compositionality – in particular, by raising the prospect that compositionality can be understood as an emergent property of an inner-loop learning algorithm. We elaborate on this hypothesis and consider its empirical predictions regarding the neural mechanisms and development of human compositionality.

Meta-learned models of cognition make optimal predictions for the actual stimuli presented to participants, but investigating judgment biases by constraining neural networks will be unwieldy. We suggest combining them with cognitive process models, which are more intuitive and explain biases. Rational process models, those that can sequentially sample from the posterior distributions produced by meta-learned models, seem a natural fit.

We propose that a principled understanding of meta-learning, as aimed for by the authors, benefits from linking the focus on learning with an equally strong focus on structure, which means to address the question: What are the meta-structures that can guide meta-learning?

Building on the affectivism approach, we expand on Binz et al.'s meta-learning research program by highlighting that emotion and other affective phenomena should be key to the modeling of human learning. We illustrate the added value of affective processes for models of learning across multiple domains with a focus on reinforcement learning, knowledge acquisition, and social learning.

Binz et al. argue that meta-learned models offer a new paradigm to study human cognition. Meta-learned models are proposed as alternatives to Bayesian models based on their capability to learn identical posterior predictive distributions. In our commentary, we highlight several arguments that reach beyond a predictive distribution-based comparison, offering new perspectives to evaluate the advantages of these modeling paradigms.

Meta-learning offers a promising framework to make sense of some parts of decision-making that have eluded satisfactory explanation. Here, we connect this research to work in animal behaviour and cognition in order to shed light on how and whether meta-learning could help us to understand the evolution of cognition.

We argue that the type of meta-learning proposed by Binz et al. generates models with low interpretability and falsifiability that have limited usefulness for neuroscience research. An alternative approach to meta-learning based on hyperparameter optimization obviates these concerns and can generate empirically testable hypotheses of biological computations.

This commentary examines the synergy between meta-learned models of cognition and integrative learning in enhancing animal and human learning outcomes. It highlights three integrative learning modes – holistic integration of parts, top-down reasoning, and generalization with in-depth analysis – and their alignment with meta-learned models of cognition. This convergence promises significant advances in educational practices, artificial intelligence, and cognitive neuroscience, offering a novel perspective on learning and cognition.

We are encouraged by the many positive commentaries on our target article. In this response, we recapitulate some of the points raised and identify synergies between them. We have arranged our response based on the tension between data and architecture that arises in the meta-learning framework. We additionally provide a short discussion that touches upon connections to foundation models.

One of the most remarkable manifestations of social cohesion in large-scale entities is the belief in a shared, distinct, and ancestral past. Human communities around the world take pride in their ancestral roots, commemorate their long history of shared experiences, and celebrate the distinctiveness of their historical trajectory. Why do humans put so much effort into celebrating a long-gone past? Integrating insights from evolutionary psychology, social psychology, evolutionary anthropology, political science, cultural history, and political economy, we show that the cultural success of historical myths is driven by a specific adaptive challenge for humans: The need to recruit coalitional support to engage in large-scale collective action and prevail in conflicts. By showcasing a long history of cooperation and shared experiences, these myths serve as super-stimuli, activating specific features of social cognition and drawing attention to cues of fitness interdependence. In this account, historical myths can spread within a population without requiring group-level selection, as long as individuals have a vested interest in their propagation and strong psychological motivations to create them. Finally, this framework explains not only the design features of historical myths, but also important patterns in their cross-cultural prevalence, interindividual distribution, and particular content.

The “historical myths” addressed in the target article are but one type of societal meta-narrative, a cognitive framework for understanding the story of one's group: Its origins, purpose, turning points, threats and opportunities, key relationships, and the appropriate affect for group members. Engagement with the broader literature on meta-narratives, including political and sacred myths, and on group entitativity is recommended.

Myths about a remote shared past can certainly promote cooperation between distantly related people, seemingly via their impact on our social cognition, and ultimately facilitate the achievement of complex tasks in large-scale societies. Nonetheless, the creation and transmission of these complex narratives are not possible without the parallel development of sophisticated language(s), endowed with properties like displacement (enabling mental travels in space and time) and complex syntax (enabling the assembly and communication of complex thoughts).

We propose that historical myths fall into two distinctive categories: Traumatic and cooperative. Traumatic myths, highlighting collective suffering, can undermine trust and foster conspiracy theories, whereas cooperative myths, emphasizing collective action, enhance group cohesion and within-group coalition building. Psychological and sociological evidence supports these divergent impacts of historical myths both in nations and social movements.

Sijilmassi et al. claim that historical myths are technologies of recruitment that mimic cues of fitness interdependence. Paradoxically, they also claim that people are vigilant and that these myths might not and do not have to convince others, which raises questions about how these myths become culturally successful. Thinking about historical myths as commitment devices helps overcome this paradox.

The explanatory model proposed by Sijilmassi et al. appeals to fitness interdependence, and is highly plausible for small-scale societies. We argue that it is less so in the context of the larger societies that much of their empirical evidence is drawn from, and that this is because fitness interdependence does not readily scale up in the way the model requires.

Sijilmassi et al. suggest that group myths explaining the shared history of a people succeed and propagate by leveraging cognitive cues from fitness interdependence. We offer an alternative and mutually compatible account rooting the success of group myths in cues from a different cognitive domain: The development of self-concepts.

Drawing on developmental psychopathology and thinking about the we-mode of social cognition, we propose that historical myths – be they on the scale of the family, the nation, or an ethnic group – are an expression and function of our need to join with other minds. As such, historical myths are one cognitive technology used to facilitate social learning, the transmission of culture and the relational mentalizing that underpins social and emotional functioning.

This comment seeks to extend the authors' argument by considering how perceived fitness interdependence is generated in different settings. Based primarily on research from political science, it argues that strategic agents may seek to design myths that emphasize not only the longevity of their coalitions, but also internal features such as material and status equality and institutional impartiality.

We agree with Sijilmassi et al. that historical myths are a tool for coalition recruitment. We argue, however, that a close fit between an evolved entity and an identified function does not imply that the latter is the critical evolutionary trigger. We also propose an alternative individual-centric explanation: Historical myths reduce uncertainty by providing cognitive and behavioral guidance.