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.

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 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.

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.

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.

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.

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?

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.

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.

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.

Deep meta-learning is the driving force behind advances in contemporary AI research, and a promising theory of flexible cognition in natural intelligence. We agree with Binz et al. that many supposedly “model-based” behaviours may be better explained by meta-learning than by classical models. We argue that this invites us to revisit our neural theories of problem solving and goal-directed planning.

Binz et al. propose meta-learning as a promising avenue for modelling human cognition. They provide an in-depth reflection on the advantages of meta-learning over other computational models of cognition, including a sound discussion on how their proposal can accommodate neuroscientific insights. We argue that active inference presents similar computational advantages while offering greater mechanistic explanatory power and biological plausibility.

We summarize the recent progress made by probabilistic programming as a unifying formalism for the probabilistic, symbolic, and data-driven aspects of human cognition. We highlight differences with meta-learning in flexibility, statistical assumptions and inferences about cogniton. We suggest that the meta-learning approach could be further strengthened by considering Connectionist and Bayesian approaches, rather than exclusively one or the other.

Binz et al. argue that meta-learned models are essential tools for understanding adult cognition. Here, we propose that these models are particularly useful for testing hypotheses about why learning processes change across development. By leveraging their ability to discover optimal algorithms and account for capacity limitations, researchers can use these models to test competing theories of developmental change in learning.

I propose that meta-learned models, and in particular the situation-aware deployment of “learning-to-infer” modules can be advantageously extended to domains commonly thought to lie outside the cognitive, such as motivations and preferences on one hand, and the effectuation of micro- and coping-type behaviors.

Meta-learning is even more broadly relevant to the study of inductive biases than Binz et al. suggest: Its implications go beyond the extensions to rational analysis that they discuss. One noteworthy example is that meta-learning can act as a bridge between the vector representations of neural networks and the symbolic hypothesis spaces used in many Bayesian models.

Abundant experimental evidence illustrates violations of Bayesian models across various cognitive processes. Quantum cognition capitalizes on the limitations of Bayesian models, providing a compelling alternative. We suggest that a generalized quantum approach in meta-learning is simultaneously more robust and flexible, as it retains all the advantages of the Bayesian framework while avoiding its limitations.

The authors' aim is to build “more biologically plausible learning algorithms” that work in naturalistic environments. Given that, first, human learning and memory are inextricable, and, second, that much human learning is unconscious, can the authors' first research question of how people improve their learning abilities over time be answered without addressing these two issues? I argue that it cannot.

The implementation of meta-learning targeted by Binz et al. inherits benefits and drawbacks from its nature as a connectionist model. Drawing from historical debates around bottom-up and top-down approaches to modeling in cognitive science, we should continue to bridge levels of analysis by constraining meta-learning and meta-learned models with complementary evidence from across the cognitive and computational sciences.

Binz et al. propose a general framework for meta-learning and contrast it with built-by-hand Bayesian models. We comment on some architectural assumptions of the approach, its relation to the active inference framework, its potential applicability to living systems in general, and the advantages of the latter in addressing the explanation problem.