The best game in town: The reemergence of the language-of-thought hypothesis across the cognitive sciences

Behavioral and Brain Sciences, Journal Year: 2022, Volume and Issue: 46

Published: Dec. 6, 2022

Mental representations remain the central posits of psychology after many decades scrutiny. However, there is no consensus about representational format(s) biological cognition. This paper provides a survey evidence from computational cognitive psychology, perceptual developmental comparative and social concludes that one type format routinely crops up language-of-thought (LoT). We outline six core properties LoTs: (i) discrete constituents; (ii) role-filler independence; (iii) predicate-argument structure; (iv) logical operators; (v) inferential promiscuity; (vi) abstract content. These cluster together throughout science. Bayesian modeling, compositional features object perception, complex infant animal reasoning, automatic, intuitive cognition in adults all implicate LoT-like structures. Instead regarding LoT as relic previous century, researchers science philosophy-of-mind must take seriously explanatory breadth LoT-based architectures. grant mind may harbor formats architectures, including iconic associative structures well deep-neural-network-like computational/representational approaches to continue advance, classical symbolic - is, LoTs only prove more flexible well-supported over time.

Language: Английский

---

On Logical Inference over Brains, Behaviour, and Artificial Neural Networks

Computational Brain & Behavior, Journal Year: 2023, Volume and Issue: 6(2), P. 213 - 227

Published: Feb. 13, 2023

Abstract In the cognitive, computational, and neuro-sciences, practitioners often reason about what computational models represent or learn, as well algorithm is instantiated. The putative goal of such reasoning to generalize claims model in question, mind brain, neurocognitive capacities those systems. Such inference based on a model’s performance task, whether that approximates human behavior brain activity. Here we demonstrate how argumentation problematizes relationship between their targets; place emphasis artificial neural networks (ANNs), though any theory-brain falls into same schema at risk. this paper, inferences from ANNs brains back within formal framework — metatheoretical calculus order initiate dialogue both are broadly understood used, best formally characterize them functions. To these ends, express published record models’ successes failures first-order logic. Our proposed formalization describes decision-making processes enacted by scientists adjudicate over theories. We formalizing literature can uncover potential deep issues theory related phenomena. discuss means for research cognitive science, neuroscience, psychology; it when they lose ability mediate data meaningful way; our fields deploy performing high-level scientific inference.

Language: Английский

---

Toward a neuroscience of natural behavior

Current Opinion in Neurobiology, Journal Year: 2024, Volume and Issue: 86, P. 102859 - 102859

Published: April 6, 2024

One of the most exciting new developments in systems neuroscience is progress being made toward neurophysiological experiments that move beyond simplified laboratory settings and address richness natural behavior. This enabled by technological advances such as wireless recording freely moving animals, automated quantification behavior, methods for analyzing large data sets. Beyond empirical data, however, there also a need theories concepts to interpret data. Such particular challenges which often differ significantly from scenarios studied traditional settings. Here, we discuss some strategies developing novel example hypotheses proposed.

Language: Английский

---

Successes and critical failures of neural networks in capturing human-like speech recognition

Neural Networks, Journal Year: 2023, Volume and Issue: 162, P. 199 - 211

Published: Feb. 25, 2023

Natural and artificial audition can in principle acquire different solutions to a given problem. The constraints of the task, however, nudge cognitive science engineering qualitatively converge, suggesting that closer mutual examination would potentially enrich hearing systems process models mind brain. Speech recognition — an area ripe for such exploration is inherently robust humans number transformations at various spectrotemporal granularities. To what extent are these robustness profiles accounted by high-performing neural network systems? We bring together experiments speech under single synthesis framework evaluate state-of-the-art networks as stimulus-computable, optimized observers. In series experiments, we (1) clarify how influential manipulations literature relate each other natural speech, (2) show granularities which machines exhibit out-of-distribution robustness, reproducing classical perceptual phenomena humans, (3) identify specific conditions where model predictions human performance differ, (4) demonstrate crucial failure all perceptually recover do, alternative directions theory building. These findings encourage tighter synergy between audition.

Language: Английский

---

Reclaiming AI as a Theoretical Tool for Cognitive Science

Computational Brain & Behavior, Journal Year: 2024, Volume and Issue: unknown

Published: Sept. 27, 2024

Language: Английский

---

Gestalt theory: A revolution put on pause? Prospects for a paradigm shift in the psychological sciences

New Ideas in Psychology, Journal Year: 2023, Volume and Issue: 71, P. 101036 - 101036

Published: July 10, 2023

Language: Английский

---

CBGTPy: An extensible cortico-basal ganglia-thalamic framework for modeling biological decision making

PLoS ONE, Journal Year: 2025, Volume and Issue: 20(1), P. e0310367 - e0310367

Published: Jan. 14, 2025

Here we introduce CBGTPy, a virtual environment for designing and testing goal-directed agents with internal dynamics that are modeled on the cortico-basal-ganglia-thalamic (CBGT) pathways in mammalian brain. CBGTPy enables researchers to investigate of CBGT system during variety tasks, allowing formation testable predictions about animal behavior neural activity. The framework has been designed around principle flexibility, such many experimental parameters decision making paradigm can be easily defined modified. demonstrate capabilities across range single multi-choice highlighting ease set up biologically realistic it produces. We show is extensible enough apply protocols allow implementation model extensions minimal developmental effort.

Language: Английский

---

Dissecting the effectiveness of deep features as metric of perceptual image quality

Neural Networks, Journal Year: 2025, Volume and Issue: 185, P. 107189 - 107189

Published: Jan. 27, 2025

Language: Английский

---

Coordinating multiple mental faculties during learning

Scientific Reports, Journal Year: 2025, Volume and Issue: 15(1)

Published: Feb. 13, 2025

Abstract Complex behavior is supported by the coordination of multiple brain regions. How do regions coordinate absent a homunculus? We propose achieved controller-peripheral architecture in which peripherals (e.g., ventral visual stream) aim to supply needed inputs their controllers hippocampus and prefrontal cortex) while expending minimal resources. developed formal model within this framework address how support rapid learning from few example images. The captured higher-level activity controller shaped lower-level representations, affecting precision sparsity manner that paralleled measures. In particular, peripheral encoded information extent smooth operation controller. Alternative models optimized gradient descent irrespective architectural constraints could not account for human or responses, and, typical standard deep approaches, were unstable trial-by-trial learners. While previous work offered accounts specific faculties, such as perception, attention, learning, approach step toward addressing next generation questions concerning faculties coordinate.

Language: Английский

---

One more trip to Barcetona: on the special status of visual similarity effects in city names

Psychological Research, Journal Year: 2023, Volume and Issue: 88(1), P. 271 - 283

Published: June 23, 2023

Previous research has shown that, unlike misspelled common words, brand names are sensitive to visual letter similarity effects (e.g., amazom is often recognized as a legitimate name, but not amazot). This pattern poses problems for those models that assume word identification exclusively based on abstract codes. Here, we investigated the role of using another type presented in more homogenous format than words: city names. We found effect Barcetona was word, Barcesona) relatively short durations stimuli (200 ms; Experiment 2), when were until response (Experiment 1). Notably, words did show brief 200- and 150-ms votume vosume; Experiments 3-4). These findings provide further evidence consistency presentations may shape representation mental lexicon, which be salient scenarios where processing resources limited exposure presentations).

Language: Английский

---