Editorial: AI and neuroscience: integrating knowledge, reasoning, and theory of mind.

Editorial: AI and neuroscience: integrating knowledge, reasoning, and theory of mind - PMC Skip to main content An official website of the United States government Here's how you know Here's how you know Search Log in Primary site navigation Search Log in Search PMC Full-Text Archive Search in PMC Journal List User Guide As a library, NLM provides access to scientific literature. Inclusion in an NLM database does not imply endorsement of, or agreement with, the contents by NLM or the National Institutes of Health. Learn more: PMC Disclaimer | PMC Copyright Notice editorial Front Comput Neurosci . 2026 May 19;20:1859797. doi: 10.3389/fncom.2026.1859797 Editorial: AI and neuroscience: integrating knowledge, reasoning, and theory of mind Darren J Edwards 1, * , Bochao Zou 2 , Rob Lowe 3 , Andrew Owens 4 Author information Article notes Copyright and License information 1 Department of Public Health, Swansea University, Swansea, United Kingdom 2 School of Computer and Communication Engineering, University of Science and Technology, Beijing, China 3 School of Psychology, Swansea University, Swansea, United Kingdom 4 Jaguar Land Rover, Warwick, United Kingdom * Correspondence: Darren J. Edwards, d.j.edwards@swansea.ac.uk Roles Darren J Edwards : Writing – original draft, Writing – review & editing Bochao Zou : Writing – original draft, Writing – review & editing Rob Lowe : Writing – review & editing Andrew Owens : Writing – original draft, Writing – review & editing Received 2026 Apr 19; Accepted 2026 Apr 22; Collection date 2026. Keywords: artificial intelligence, cognition, computation, neuroscience, reasoning Copyright © 2026 Edwards, Zou, Lowe and Owens. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY) . The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms. PMC Copyright notice PMCID: PMC13226194  PMID: 42238304 Artificial intelligence and neuroscience have been converging on a shared problem about how to explain and ultimately build systems (biological or synthetic) that can acquire knowledge from experience, reason under uncertainty, and coordinate perspectives from self-model representations to other minds ( Hassabis et al., 2017 ; Langley et al., 2022 ; Limanowski and Blankenburg, 2013 ; Nawaz et al., 2025 ). The “ AI and neuroscience: integrating knowledge, reasoning, and theory of mind ” theme captures this convergence by explicitly highlighting a broad body of research linking accounts of neural information processing with computational architectures that can learn, generalize, and remain interpretable. At a broad level, the contributions in this Research Topic can be read as collectively operating across three complementary levels. First, they investigate biological substrates of computation and the representational constraints that come with real neural tissue. Second, they advance architectures and modeling frameworks that treat cognition as an evolving repertoire of learned competencies rather than a set of isolated tasks. Third, they address human–AI coupling, i.e., how AI systems can extend cognition without displacing the very internal knowledge structures that make reasoning and perspective-taking possible in the first place. A summary of the article contributions Edwards extends the functional contextual N -Frame model ( Edwards, 2023 , 2024 ) as an integrative theory linking predictive coding ( Friston, 2018 ; Spratling, 2017 ), QBism-style observer dependence ( Fuchs et al., 2014 ), and evolutionary dynamics to explain how belief updating and decision-making in humans (and potentially AI) can be modeled within a quantum-cognitive functional contextual formalism. The paper positions consciousness as an active, context-sensitive participant in “actualization” (bridging quantum potentiality to classical outcomes), uses cognitive fallacies (e.g., conjunction effects) as structured signatures of quantum context-dependent inference rather than mere cognitive classical irrationality, and proposes testable boundaries/parameters relevant to AI consciousness beyond standard AI benchmark performance. In the Research Topic's terms, it offers a unifying scaffold for knowledge (context-shaped representations), reasoning (state updates under constraints), and the foundations needed for theory of mind (self-referential observer modeling), while explicitly connecting these claims to experimentally oriented predictions and formal operator-style descriptions, in order to potentially develop safe and ethical AI. Tütüncü and Gonzalez-Franco argues that “algorithmic suffering” is best treated as a comparative neuroscientific lens rather than as a claim that machines literally feel pain. They identify parallels between human and AI cognition in reward prediction error, Bayesian belief updating, and risk anticipation, while emphasizing a decisive asymmetry: in humans, prediction errors can become globally integrated into a conscious self-model and experienced as threats to meaning and integrity, whereas in current AI they remain numerical updates without phenomenology. Framed within this Research Topic, the paper helps clarify where AI's performance (its observable behavior and outputs) may mimic cognition while still falling short of the selfhood, meaning, and experiential integration that would be required for genuinely human mind-like understanding. Küchler et al. provide a clear “wetware-to-logic” bridge by demonstrating that engineered in vitro neuronal networks with controlled topology can implement basic Boolean operations (including NAND/OR) using stimulation and readout on high-density microelectrode arrays. Importantly, the work goes beyond a proof-of-principle gate: it interrogates encoding and decoding choices (rate-based vs. spike-timing-based schemes such as TTFS), highlighting how representational format constrains how information can be reliably encoded, transmitted, and thus computed. Such minimal networks may serve as building blocks for hybrid intelligence and biocomputing. In the context of this Research Topic, the paper grounds “knowledge and reasoning” in the physical realities of neural signaling and frames neural computation as an input–output mapping that can be characterized and engineered. Chowdhury et al. review the rapidly developing landscape of inner speech recognition (ISR), i.e., the decoding of covert/inner speech from neural signals, and position machine learning as the primary driver of progress across the ISR pipeline (signal acquisition, preprocessing, feature extraction, and classification). The review synthesizes cognitive models of inner speech alongside a comparative analysis of machine learning (ML) and deep learning (DL) approaches [e.g., support vector machines (SVMs) and random forests vs. convolutional neural network (CNN-based models)], while emphasizing persistent barriers such as signal-to-noise limitations, inter-subject variability, interpretability, and challenges for real-time deployment. Conceptually, inner speech sits at the intersection of language, self-regulation, and planning; as such, ISR research provides a natural bridge toward higher-order cognitive functions (i.e., neural signals → inner speech decoding → internal cognition). In the context of this Research Topic, the paper can be read as exemplifying the “integrating knowledge and reasoning” theme as an applied neuroscience-to-AI translation problem with significant clinical and assistive-technology implications. Prudkov advances a hypothesis-and-theory account of AGI grounded in the “goals–means correspondence,” arguing that the central challenge for general intelligence is not merely producing competent outputs, but establishing and dynamically maintaining the correspondence between goals and the means available in an evolving world. The paper critiques two conventional agent architectures in the form of those with jointly specified goals and means at “birth” (that is, at the moment the agent is created or initialized), and those in which goals and means are constructed separately. It proposes an alternative: the joint construction of arbitrary goals and means under a criterion of minimal construction cost, framed as a cognitive analog of least action. This proposal is explicitly developmental (agents “grow” by altering structure), and it connects naturally to neuroscience-informed views of planning and prefrontal control, while supplying a formal lens on how reasoning can emerge as a type of goal-directed process. Within this Research Topic, it can be read as contributing a unifying architectural principle for linking learning, reasoning, and adaptive agency. O'Sullivan et al. introduce “Affinity,” a visual analytics tool designed to make computational models of stimulus equivalence and derived relational responding more transparent and experimentally useful. Built on Enhanced Equivalence Projective Simulation, the tool provides real-time visualizations of the evolving relational memory of an agent and operationalizes Relational Density Theory ( Belisle and Dixon, 2020 ) by modeling higher-order network properties (e.g., density, volume, and mass) that may explain resistance to change and the dynamics of relational learning. As a contribution to “knowledge and reasoning,” this work is significant because relational generalization and abstraction are foundational to language-like cognition. More broadly, the framework connects computational modeling of relational learning with cognitive and behavioral science accounts of flexible, context-sensitive responding. Affinity's (their visual analytics tool) emphasis on interpretability and experiment-as-interface also aligns with a wider shift toward explainable AI, where models are designed not only to perform but to reveal their internal process dynamics in ways that can be systematically analyzed and compared to empirical data. Klein and Klein broaden the scope from building intelligent systems to preserving and redesigning environments to support human cognition in AI-rich contexts. Their “extended hollowed mind” framework argues that generative AI produces a dual outcome in education: it can extend cognition while also enabling a “cognitive bypass” that weakens the internal structures required for deep learning and evaluative judgment, leading to a lack of deep cognitive understanding. They introduce the term “Sovereignty Trap” (whereby humans have the tendency to cede judgment to an authoritative system and trust AI too much) and reframe the educational target as a “Fortified Mind,” defined as an internal architecture of knowledge and metacognitive skills necessary for cognitive sovereignty (whereby humans develop solid knowledge and critical thinking skills). The paper's relevance to this Research Topic is 2-fold: it anchors the discussion in cognitive science and neurobiological constraints on effortful reasoning, and it treats reasoning capacity as an internal architecture that must be actively trained. It therefore functions as a crucial boundary condition on the design of AI systems for learning, i.e., tools that replace human reasoning risk degrading the very cognitive competencies required to interpret, validate, and effectively use them. Finally, Johansson and Hammer review the Machine-Psychology program implemented in the Non-Axiomatic Reasoning System (NARS), presenting a staged developmental roadmap from operant learning to abstraction, functional equivalence, and arbitrarily applicable relational responding. A central strength of this work is its explicit mapping between architectural mechanisms (e.g., temporal inference, resource limitations, relational generalization) and well-established psychological phenomena, treating cognition as a learnable process rather than as a collection of task-specific solutions. The manuscript also links relational learning and contextual control to perspective-taking and theory-of-mind–related abilities, situating cognitive architecture as a bridge between neuroscience-relevant process models and AGI ambitions. In the context of this Research Topic, it directly embodies the integrative thesis: knowledge emerges through experience, reasoning operates under bounded resources, and higher social-cognitive competencies can be built from tractable learning primitives. In conclusion, taken together, these contributions sketch a coherent multi-scale picture of integration. At the substrate level, engineered neuronal networks clarify what computation in biological substrates can look like, and why representational choices (e.g., rate vs. timing, i.e., how often neurons fire, such as 50 spikes per second vs. when they fire, such as precise millisecond patterns, order, synchrony) matter for reliable inference. At the systems level, architectural proposals and interpretable modeling tools advance a view of intelligence as developmental: learned competencies accumulate from simple feedback-driven adaptation to symbolic and relational generalization. At the human interface level, the educational and epistemic framing reminds us that theory-of-mind-related, reasoning-enabled AI will only be beneficial if humans retain the internal knowledge structures required to evaluate, contextualize, and govern it. From this Research Topic, a shared message emerges. Societal progress and wellbeing in the age of AI clearly come less from chasing isolated benchmarks and more from building process bridges: between neural codes and computation, between learning histories and relational abstraction, and between tool design and the preservation of human cognitive sovereignty. This is precisely the integration challenge at the heart of “ AI and neuroscience: integrating knowledge, reasoning, and theory of mind ” and we hope it will lead to a new generation of intelligent systems designed not only for performance but for the advancement of a fairer, healthier, and more resilient society. Editorial on the Research Topic AI and neuroscience: integrating knowledge, reasoning, and theory of mind Footnotes Edited and reviewed by: Si Wu , Peking University, China Author contributions DE: Writing – original draft, Writing – review & editing. BZ: Writing – original draft, Writing – review & editing. RL: Writing – review & editing. AO: Writing – original draft, Writing – review & editing. Conflict of interest AO was employed by the Jaguar Land Rover. The remaining author(s) declared that this work was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest. Generative AI statement The author(s) declared that generative AI was not used in the creation of this manuscript. Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us. Publisher's note All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher. References Belisle J., Dixon M. R. (2020). Relational density theory: nonlinearity of equivalence relating examined through higher-order volumetric-mass-density. Perspect. Behav. Sci. 43, 259–283. doi: 10.1007/s40614-020-00248-w [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Edwards D. J. (2023). Functional contextual implementation of an evolutionary, entropy-based, and embodied free energy framework: utilizing Lagrangian mechanics and evolutionary game theory's truth vs. fitness test of the veridicality of phenomenological experience. Front. Psychol. 14:1150743. doi: 10.3389/fpsyg.2023.1150743 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Edwards D. J. (2024). A functional contextual, observer-centric, quantum mechanical, and neuro-symbolic approach to solving the alignment problem of artificial general intelligence: safe AI through intersecting computational psychological neuroscience and LLM architecture for emergent theory of mind. Front. Comput. Neurosci. 18:1395901. doi: 10.3389/fncom.2024.1395901 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Friston K. (2018). Does predictive coding have a future? Nat. Neurosci. 21, 1019–1021. doi: 10.1038/s41593-018-0200-7 [ DOI ] [ PubMed ] [ Google Scholar ] Fuchs C. A., Mermin N. D., Schack R. (2014). An introduction to QBism with an application to the locality of quantum mechanics. Am. J. Phys. 82, 749–754. doi: 10.1119/1.4874855 [ DOI ] [ Google Scholar ] Hassabis D., Kumaran D., Summerfield C., Botvinick M. (2017). Neuroscience-inspired artificial intelligence. Neuron 95, 245–258. doi: 10.1016/j.neuron.2017.06.011 [ DOI ] [ PubMed ] [ Google Scholar ] Langley C., Cirstea B. I., Cuzzolin F., Sahakian B. J. (2022). Theory of mind and preference learning at the interface of cognitive science, neuroscience, and AI: a review. Front. Artif. Intell. 5:778852. doi: 10.3389/frai.2022.778852 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Limanowski J., Blankenburg F. (2013). Minimal self-models and the free energy principle. Front. Hum. Neurosci. 7:547. doi: 10.3389/fnhum.2013.00547 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Nawaz U., Anees-ur-Rahaman M., Saeed Z. (2025). A review of neuro-symbolic AI integrating reasoning and learning for advanced cognitive systems. Intell. Syst. Applic. 26:200541. doi: 10.1016/j.iswa.2025.200541 [ DOI ] [ Google Scholar ] Spratling M. W. (2017). A review of predictive coding algorithms. Brain Cogn. 112, 92–97. doi: 10.1016/j.bandc.2015.11.003 [ DOI ] [ PubMed ] [ Google Scholar ] Articles from Frontiers in Computational Neuroscience are provided here courtesy of Frontiers Media SA ACTIONS View on publisher site PDF (113.9 KB) Cite Collections Permalink RESOURCES Similar articles Cited by other articles Links to NCBI Databases Follow NCBI NCBI on X (formerly known as Twitter) NCBI on Facebook NCBI on LinkedIn NCBI on GitHub NCBI RSS feed Connect with NLM NLM on X (formerly known as Twitter) NLM on Facebook NLM on YouTube National Library of Medicine 8600 Rockville Pike Bethesda, MD 20894 Web Policies FOIA HHS Vulnerability Disclosure Help Accessibility Careers NLM NIH HHS USA.gov Back to Top

Related documents

• Cognitive Neuroscience#3318
• Mechanisms, Coherence, and Theory Choice in the Cognitive Neurosciences#739018
• NEUROR: Exploring the Neural Basis of Reasoning#836002
• “Surgery is Thinking”: cognitive neuroscience perspective for the AI age.#25658
• An fMRI Exploration of Information Processing in Electronic Networks of Practice#387068