Psychologists and neuroscientists extensively rely on computational models for studying and analyzing the human mind. Traditionally, such computational models have been hand-designed by expert researchers. Two prominent examples are cognitive architectures and Bayesian models of cognition. Although the former requires the specification of a fixed set of computational structures and a definition of how these structures interact with each other, the latter necessitates the commitment to a particular prior and a likelihood function that – in combination with Bayes' rule – (...) determine the model's behavior. In recent years, a new framework has established itself as a promising tool for building models of human cognition: the framework of meta-learning. In contrast to the previously mentioned model classes, meta-learned models acquire their inductive biases from experience, that is, by repeatedly interacting with an environment. However, a coherent research program around meta-learned models of cognition is still missing to date. The purpose of this article is to synthesize previous work in this field and establish such a research program. We accomplish this by pointing out that meta-learning can be used to construct Bayes-optimal learning algorithms, allowing us to draw strong connections to the rational analysis of cognition. We then discuss several advantages of the meta-learning framework over traditional methods and reexamine prior work in the context of these new insights. (shrink)

Meta-learning is learning to learn, which includes the development of capacities to transfer what people learned in one specific domain to other domains. It facilitates finetuning learning parameters and setting priors for effective and optimal learning in novel contexts and situations. Recent advances in research on artificial intelligence have reported meta-learning is essential in improving and optimizing the performance of trained models across different domains. In this paper, I suggest that meta-learning (...) plays fundamental roles in practical wisdom and its cultivation within moral domains. I support my argument with findings from studies in psychology, neuroscience, and computer science. Moreover, I suggested concrete ideas for moral education to foster wisdom cultivation along with meta-learning with evidence from studies across various fields. Particularly, I underscored that exemplars, including role models and mentors, play fundamental roles in education. Finally, I discussed the potential implications of the introduction of the concept, meta-learning, to moral education addressing concerns and issues related to socio-cultural diversity. (shrink)

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.

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.

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.

The meta-learning framework proposed by Binz et al. would gain significantly from the inclusion of affective and homeostatic elements, currently neglected in their work. These components are crucial as cognition as we know it is profoundly influenced by affective states, which arise as intricate forms of homeostatic regulation in living bodies.

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.

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 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.

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.

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.

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.

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.

While there is general consensus that robust forms of social learning enable the possibility of human cultural evolution, the specific nature, origins, and development of such learning mechanisms remains an open issue. The current paper offers an action-based approach to the study of social learning in general and imitation learning in particular. From this action-based perspective, imitation itself undergoes learning and development and is modeled as an instance of social meta-learning – children (...) class='Hi'>learning how to use others as a resource for further learning. This social meta-learning perspective is then applied empirically to an ongoing debate about the reason children imitate causally unnecessary actions while learning about a new artifact (i.e., over-imitate). Results suggest that children over-imitate because it is the nature of learning about social realities in which cultural artifacts are a central aspect. (shrink)

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?

We propose extending business ethics education beyond the formal curriculum to the hidden curriculum where messages about ethics and values are implicitly sent and received. In this meta-learning approach, students learn by becoming active participants in an honorable business school community where real ethical issues are openly discussed and acted upon. When combined with formal ethics instruction, this meta-learning approach provides a framework for a proposed comprehensive program of business ethics education.

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.

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.

Quantum active Bayesian inference and quantum Markov blankets enable robust modeling and simulation of difficult-to-render natural agent-based classical inferential paradoxes interfaced with task-specific environments. Within a non-realist cognitive completeness regime, quantum Markov blankets ensure meta-learned irrational decision making is fitted to explainable manifolds at optimal free energy, where acceptable incompatible observations or temporal Bell-inequality violations represent important verifiable real-world outcomes.

We challenge Binz et al.'s claim of meta-learned model superiority over Bayesian inference for large world problems. While comparing Bayesian priors to model-training decisions, we question meta-learning feature exclusivity. We assert no special justification for rational Bayesian solutions to large world problems, advocating exploring diverse theoretical frameworks beyond rational analysis of cognition for research advancement.

In this chapter we, as scholars of authoritarianism, discuss the ‘red lines’, a term used in authoritarian contexts to denote topics that are highly politically sensitive. We first describe commonalities in what the red lines are in different contexts, distinguishing between hard red lines and more fluid ones. We describe how we navigate red lines in fieldwork by offering a depoliticized, but not untrue, version of our research; how we adapt our wording and behavior to remain within the red lines, (...) but still give us meaningful research results; and how we respond when the red lines shift, and words and behaviors previously acceptable become taboo, or vice versa. (shrink)

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.

Objective: Mindfulness-based intervention has been proposed to alleviate loneliness and improve social connectedness. Several randomized controlled trials have been conducted to evaluate the effectiveness of MBI. This study aimed to critically evaluate and determine the effectiveness and safety of MBI in alleviating the feeling of loneliness.Methods: We searched Medline, Embase, PsycInfo, Cochrane CENTRAL, and AMED for publications from inception to May 2020. We included RCTs with human subjects who were enrolled in MBI with loneliness as an outcome. The quality of (...) evidence was assessed using Cochrane's Risk of Bias tool and Grading of Recommendations Assessment, Development, and Evaluation. A random-effects model was used for meta-analysis.Results: Out of 92 articles identified, eight studies involving 815 participants were included in this study. Most trials conducted a minimum of 8 weeks of MBI. Most of the trials used UCLA-Loneliness Scale. A pooled analysis combining three trials and compared with wait-list showed significant improvement in loneliness score reduction using the UCLA-R scale with MD of −6.33 [95% confidence interval : −9.39, −3.26]. Subgroup analysis with only two Cognitively-Based Compassion Training trials also showed similar MD of −6.05. The overall quality of evidence was low.Conclusions: Mindfulness intervention with an average length of 8-week duration significantly improved the population's loneliness level with no mental health issue. However, this evidence had a low GRADE level. (shrink)

BackgroundThe advent of learning healthcare systems (LHSs) raises an important implementation challenge concerning how to request and manage consent to support secondary use of data in learning cycles, particularly research activities. Current consent models in Quebec were not established with the context of LHSs in mind and do not support the agility and transparency required to obtain consent from all involved, especially the citizens. Therefore, a new approach to consent is needed. Previous work identified the meta-consent model (...) as a promising alternative to fulfill the requirements of LHSs, particularly large-scale deployments. We elicited the public’s attitude toward the meta-consent model to evaluate if the model could be understood by the citizens and would be deemed acceptable to prepare for its possible implementation in Quebec.MethodsEight focus groups, with a total of 63 members of the general public from various backgrounds were conducted in Quebec, Canada, in 2019. Explicit attention was given to literacy levels, language spoken at home and rural vs urban settings. We assessed attitudes, concerns and facilitators regarding key components of the meta-consent model: predefined categories to personalized consent requests, a dynamic web-based infrastructure to record meta-consent, and default settings. To analyse the discussions, a thematic content analysis was performed using a qualitative software.ResultsOur findings showed that participants were supportive of this new approach of consent as it promotes transparency and offers autonomy for the management of their health data. Key facilitators were identified to be considered in the implementation of a meta-consent model in the Quebec LHSs: information and transparency, awareness campaigns, development of educational tools, collaboration of front-line healthcare professionals, default settings deemed acceptable by the society as well as close partnerships with recognized and trusted institutions.ConclusionsThis qualitative study reveals the openness of a sample of the Quebec population regarding the meta-consent model for secondary use of health data for research. This first exploratory study conducted with the public is an important step in guiding decision-makers in the next phases of implementing the various strategies to support access and use of health data in Quebec. (shrink)

The meta-inductive justification of induction is usually regarded as a social learning strategy. But, pre-theoretically, induction can be justified even for an isolated thinker, incapable of, or unwilling to engage in, any social interactions. This paper presents reinforcement learning as a natural meta-inductive strategy for such isolated thinkers. The meta-inductive reinforcement learner learns to choose an optimal prediction method among the available methods by applying trial-and-error learning. We use computer simulations to show that, under (...) certain plausible assumptions, trial-and-error learning indeed leads the thinker to favor induction over non-inductive predictive methods. Under further plausible assumptions, the meta-inductive thinker is even seen to outperform all object-level methods. The results are shown to hold in both stationary and dynamic environments. (shrink)

Iowa Gambling Task (IGT) is one of the most widely used tools to assess economic decision-making. However, the research tradition on aging and the Iowa Gambling Task (IGT) has been mainly focused on the overall performance of older adults in relation to younger or clinical groups, remaining unclear whether older adults are capable of learning along the task. We conducted a meta-analysis to examine older adults’ decision-making on the IGT, to test the effects of aging on reversal (...) class='Hi'>learning (45 studies) and to provide normative data on total and block net scores (55 studies). From the accumulated empirical evidence, we found an average total net score of 7.55 (± 25.9). We also observed a significant reversal learning effect along the blocks of the IGT, indicating that older adults inhibit the prepotent response towards immediately attractive options associated with high losses, in favor of initially less attractive options associated with long-run profit. During block 1, decisions of older adults led to a negative gambling net score, reflecting the expected initial pattern of risk-taking. However, the shift towards more safe options occurred between block 2 (small-to-medium effect size) and blocks 3, 4, 5 (medium-to-large effect size). These main findings highlight that older adults are able to move from the initial uncertainty, when the possible outcomes are unknown, to decisions based on risk, when the outcomes are learned and may be used to guide future adaptive decision-making. (shrink)

This paper builds on two of James Baldwin’s insights concerning the epistemic effects of American racism on Black Americans. I argue that he anticipated important points later raised by José Medina and Charles Mills, according to which those subjected to discrimination may have an epistemic advantage over discriminators, are an advantage in terms of what Medina later called meta-lucidity. Building on another one of Baldwin’s insights, I then argue that racial discrimination (and by extension other forms of discrimination) can (...) also epistemically harm those subjected to it. In doing so, I relate this harm to the topic of epistemic injustice, and draw connections to work done on relevant alternatives in epistemology, on micro-aggressions, and on self-trust. I also consider the extent to which the epistemic advantage that may arise from being subjected to discrimination may mitigate the epistemic harm it inflicts. (shrink)

The sudden outbreak of COVID-19 has made blended learning widely accepted, followed by many studies committed to blended learning outcomes and student attitudes. Few studies have, however, focused on the summarized effect of blended learning. To complement this missing link, this study meta-analytically reviews blended learning outcomes and student attitudes by including 30 peer-reviewed journal articles and 70 effect sizes. It concludes that blended learning outcomes are significantly higher than the traditional learning outcomes (...) with a medium effect size, and learners hold significantly more positive attitudes toward blended learning than traditional learning with a medium effect size. Blended learning may be promising, and information technology scientists may focus on the development of more advanced and effective devices to improve blended learning effectiveness. (shrink)

Background: Differential learning is a motor learning method characterized by high amounts of variability during practice and is claimed to provide the learner with a higher learning rate than other methods. However, some controversy surrounds DL theory, and to date, no overview exists that compares the effects of DL to other motor learning methods.Objective: To evaluate the effectiveness of DL in comparison to other motor learning methods in the acquisition and retention phase.Design: Systematic review and (...) exploratory meta-analysis.Methods: PubMed, Web of Science, and Google Scholar were searched until February 3, 2020. To be included, studies had to be experiments where the DL group was compared to a control group engaged in a different motor learning method, studies had to describe the effects on one or more measures of performance in a skill or movement task, and the study report had to be published as a full paper in a journal or as a book chapter.Results: Twenty-seven studies encompassing 31 experiments were included. Overall heterogeneity for the acquisition phase as well as for the retention phase was large, and risk of bias was high. The meta-analysis showed an overall small effect size of 0.26 [0.10, 0.42] in the acquisition phase for participants in the DL group compared to other motor learning methods. In the retention phase, an overall medium effect size of 0.61 [0.30, 0.91] was observed for participants in the DL group compared to other motor learning methods.Discussion/Conclusion: Given the large amount of heterogeneity, limited number of studies, low sample sizes, low statistical power, possible publication bias, and high risk of bias in general, inferences about the effectiveness of DL would be premature. Even though DL shows potential to result in greater average improvements between pre- and post/retention test compared to non-variability-based motor learning methods, more high-quality research is needed before issuing such a statement. For robust comparisons on the relative effectiveness of DL to different variability-based motor learning methods, scarce and inconclusive evidence was found. (shrink)

Objectives: This study aimed to provide a comprehensive understanding of all intervention for learning burnout by meta-analyzing their effects.Methods: Relevant studies that had been published up to September 18, 2020, were identified through a systematic search of the PubMed, Web of Science, the China National Knowledge Infrastructure, and Wan Fang databases. Eligible studies included randomized control trials of any learning burnout intervention conducted among students. The Jadad scale was used to evaluate the quality of the study. Random-effect (...) meta-analyses, subgroup analyses, meta-regression analysis, and sensitivity analysis were conducted. Funnel plots and Egger's tests were used to evaluate publication bias. Duval and Tweedie's non-parametric trim-and-fill method was used to adjust the effect of publication bias.Results: Of the 5,245 articles found, 39 met the inclusion criteria for the systematic review. There were 3,400 students, including 1,847 students in the intervention group and 1,553 in the control group. A meta-analysis of 44 studies showed that the interventions were effective. Subgroup analyses were conducted according to education, scales, intervention measures, and intervention time. The results showed that, compared with the control group, the learning burnout scores of undergraduates, secondary vocational students, and middle school students were significantly lower. Based on different scales, all interventions were also effective. The funnel plot was asymmetric and consistent with the results of Egger's test. The trim-and-fill method was used, and seven missing studies were merged to obtain a symmetric funnel plot.Conclusions: This meta-analysis indicated that learning burnout interventions are effective. The subgroup analyses showed that group counseling is the most widely used, exercise intervention is probably the most effective, and 8 weeks or more is the appropriate intervention time. An integrated intervention study based on the factors of learning burnout adds value. More studies are needed to supplement the results in the future. (shrink)

In school education, teaching-learning interaction is deemed as a core process in the classroom. The fundamental neural basis underlying teaching-learning interaction is proposed to be essential for tuning learning outcomes. However, the neural basis of this process as well as the relationship between the neural dynamics and the learning outcomes are largely unclear. With non-invasive technologies such as fNIRS, hyperscanning techniques have been developed since the last decade and been applied to the field of educational neuroscience (...) for simultaneous multi-brain scanning. Hyperscanning studies suggest that the interpersonal neural synchronization during teaching-learning interaction might be an ideal neural biomarker for predicting learning outcomes. To systematically evaluate such a relationship, this meta-analysis ran on a random-effects model on 16 studies with 23 independent samples. Further moderator analyses were also performed to examine the potential influences of the style, mode, content, and the assessment method of learning outcomes. The random-effects modeling results confirmed a robust positive correlation between INS and learning outcomes. Subsequent analyses revealed that such relationship was mainly affected by both interaction style and mode. Therefore, the present meta-analysis provided a confirmatory neurocognitive foundation for teaching-learning interaction, as well as its relation to the learning outcomes, consolidated future learning and teaching studies in various disciplines including second language education with a firm methodological reference. (shrink)

Education theorists have emphasized that transformative learning is not simply a matter of students gaining access to new knowledge and information, but instead centers upon personal transformation: it alters students’ perspectives, interpretations, and responses. How should learning that brings about this sort of self-transformation be understood from the perspectives of philosophy of mind and cognitive science? Jack Mezirow has described transformative learning primarily in terms of critical reflection, meta-cognitive reasoning, and the questioning of assumptions and beliefs. (...) And within mainstream philosophy of mind, there has been a long-standing assumption that cognition and thought are brain-based, computational, disembodied processes that occur separately from emotion and affect. According to this view, selftransformation might be construed as the forging of new neural connections and the development of new cognitive “programs.” However, I will argue that the literature on embodiment and enactivism that has emerged in recent years offers us a different and more productive way to conceptualize the intended effects of transformative learning. From the standpoint of enactivism, the experience of transformative learning is thoroughly bound up with the cognitive shifts that it involves, and it also involves significant changes to the neurobiological dynamics of the living body. Moreover, personal transformation is not simply something that happens to subjects, but rather a process in which they are actively and dynamically engaged. In addition, this enactivist approach emphasizes that the learning process is fully embodied and fundamentally affective. From a phenomenological perspective, personal transformation can be understood as a pronounced alteration in cognitive-affective orientation; and from a neurobiological perspective, the development of new habits of mind can be understood as the formation of highly integrated patterns of bodily engagement and response. The upshot is that it is not just subjects’ brains that are altered over the course of transformative learning, but also their overall bodily and affective attunement to their surroundings. (shrink)

This article examines how local experiments and negotiation processes contribute to social and field-level learning. The analysis is framed within the niche development literature, which offers a framework for analyzing the relation between projects in local contexts and the transfer of local experiences into generally applicable rules. The authors examine 2 case studies drawn from a meta-analysis of 27 new energy projects. The case studies, both pertaining to biogas projects for local municipalities, illustrate the diversity of applications for (...) a technology through processes of local variation and selection. The authors examine the diversity of expectations and the negotiation and alignment of these expectations underlying the diversity of local solutions. Moreover, the authors address how the transfer of lessons from individual local experiments can follow different pathways and yet always require due attention to the social and cultural limits to the transferability of solutions. (shrink)

Evaluative learning refers to the change in the affective evaluation of a previously neutral stimulus that occurs after the stimulus has been associated with a second, positive or negative, affective stimulus. Four experiments are reported in which the AS was presented very briefly. Significant evaluative learning was observed in participants who did not notice the presentation of the affective stimuli or could not discriminate between the briefly presented positive and negative ASi when asked to do so. In two (...) other experiments, no significant learning effect was obtained. A meta-analysis performed on the present and previously reported results gave evidence for a small, though statistically reliable evaluative learning effect when ASi are presented “subliminally.” This finding supports the hypothesis that evaluative associations can be learned implicitly. (shrink)

Procedural justifications of democracy emphasize inclusiveness and respect and by doing so come into conflict with instrumental justifications that depend on voters’ competence. This conflict raises questions about jury theorems and makes their standing in democratic theory contested. We show that a type of no-regret learning called meta-induction can help to satisfy the competence assumption without excluding voters or diverse opinion leaders on an a priori basis. Meta-induction assigns weights to opinion leaders based on their past predictive (...) performance to determine the level of their inclusion in recommendations for voters. The weighting minimizes the difference between the performance of meta-induction and the best opinion leader in hindsight. The difference represents the regret of meta-induction whose minimization ensures that the recommendations are optimal in supporting voters’ competence. Meta-induction has optimal truth-tracking properties that support voters’ competence even if it is targeted by mis/disinformation and should be considered a tool for supporting democracy in hyper-plurality. (shrink)

Cognitive Science, Volume 46, Issue 9, September 2022.

Interventions that incorporated the teaching of self-regulated learning strategies are assumed to be effective in improving students' second language performance as they support students' SRL activity and self-efficacy. Nevertheless, previous meta-analyses largely focused on students' language learning achievement, while neglecting the instructional effects on their SRL strategy use and self-efficacy, two key factors in SRL models. This meta-analytic study was thus conducted to address the gap by synthesizing the evidence of SRL interventions in influencing students' L2 (...) learning achievement, strategy use, and self-efficacy. The largest effect was obtained for L2 learning achievement, followed by self-efficacy and strategy use. Moderator analysis revealed similar instructional effects on students of different age groups and education levels. The duration and intensity of intervention significantly moderated the effectiveness of SRL interventions in the L2 context, especially for strategy use and self-efficacy. The findings obtained in the current study could inform practitioners and researchers of the cumulative effects of SRL interventions in L2 classrooms and study design and student characteristics that moderate the instructional effectiveness. (shrink)