Last reviewed on October 31, 2026.
Short answer: how is AI shaping cognitive science?
AI gives cognitive scientists working models that can be tested against human behavior and brain data, and large language models are now studied as both research tools and test subjects. The influence runs the other way too: ideas from psychology and neuroscience helped inspire artificial neurons and reinforcement learning, and cognitive science supplies the methods for checking whether a system that behaves intelligently works the way people do. The similarities have limits. Today's AI systems do not learn or develop the way children do, and whether they understand anything is still debated.
AI and cognitive science grew up together. Both took shape in the 1950s around the idea that thinking can be studied as a form of information processing, and they have borrowed from each other ever since. Some of the popular claims about this relationship go too far, in both directions. This article separates what the evidence supports from what is still analogy or speculation. For the broader background, see our article on connectionism and neural networks.
The Historical Intertwining: From Turing to Transformers
The Founding Vision
Alan Turing's 1950 paper "Computing Machinery and Intelligence" proposed the imitation game, now known as the Turing test, as a practical substitute for the question "can machines think?" The 1956 Dartmouth summer workshop gave artificial intelligence its name and brought together researchers including John McCarthy, Marvin Minsky, Allen Newell and Herbert Simon. Cognitive science has its own 1956 origin story, a symposium at MIT often cited as a starting point of the cognitive revolution, and Newell and Simon presented work at both.
Their early programs had a dual purpose. The Logic Theorist (1956) proved theorems, and the General Problem Solver (1959) searched for solutions in ways compared directly with people's think-aloud reports. Newell, Shaw and Simon (1958) presented this work as a theory of human problem solving, not just as engineering.
The Cognitive Revolution and AI Winter
The cognitive revolution adopted the computer as a metaphor for the mind, with concepts such as information processing and limited-capacity memory. David Marr's three levels of analysis (computational, algorithmic and implementation) offered a shared framework for studying vision in brains and machines. Production systems, sets of if-then rules developed in the Newell and Simon tradition, became the basis both of expert systems in AI and of psychological models such as John Anderson's ACT-R.
Early symbolic AI promised more than it could deliver. Disappointing results in machine translation and elsewhere contributed to funding cuts in the 1970s, a period known as the first AI winter, and the limits of expert systems led to a second downturn around the end of the 1980s. Connectionists revived neural networks during the 1980s, and those networks are the direct ancestors of today's systems.
AI as a Laboratory for Cognitive Science
Computational Models as Theories
A computational model has to be precise in a way a verbal theory does not. To run at all, it must specify what happens at each step, which exposes hidden assumptions and produces testable predictions.
Language is the clearest current example. Schrimpf and colleagues (2021) compared many language models against human brain data and found that transformer models predicted neural responses to sentences well, and that the models best at next-word prediction fit the brain best. That supports the view that prediction plays a central role in comprehension, discussed in our article on how the brain processes language. A good statistical fit does not show that the brain uses the same mechanism, though, and in vision, deep networks trained on object recognition similarly predict activity in the primate visual system (Yamins et al., 2014).
Testing Cognitive Theories at Scale
Machine learning also lets researchers analyze data at a scale that was impractical before. Long and colleagues (2024), for example, used vision algorithms to analyze more than 37,000 line drawings made by children aged 2 to 10 at a museum kiosk, and traced how drawings carry more category information as children get older.
Models also allow experiments that are impossible with people, such as changing a system's memory or attention and seeing what follows. A related line of work, resource-rational analysis, treats human shortcuts as sensible use of limited computing power (Lieder & Griffiths, 2020). That links directly to research on why cognitive biases exist.
What AI Has Borrowed from Cognitive Science
The influence is real, but it is usually loose inspiration, not a direct copy. Hassabis and colleagues (2017) review how neuroscience has informed AI, including reinforcement learning and the replay of past experience in training. Fukushima's neocognitron (1980), an ancestor of convolutional networks, was modeled on the organization of the visual cortex.
Attention Mechanisms: From Psychology to Transformers
The word "attention" links two fields, but the shared history is thinner than it sounds. In psychology, Donald Broadbent's filter theory and the feature integration theory of Treisman and Gelade (1980) described how people select and bind information. In AI, Bahdanau and colleagues (2014) introduced attention to solve an engineering problem: translation systems struggled when a whole sentence was squeezed into one fixed-length vector, so they let the model look back at the relevant source words. The transformer (Vaswani et al., 2017) built on this and dropped recurrence. Lindsay (2020) compares attention across psychology, neuroscience and machine learning and is a useful guide to how far the parallels go. Transformers are not models of human attention.
Memory Systems: From Cognitive Architecture to Neural Networks
Psychology's distinctions between working, episodic and semantic memory offer a vocabulary that AI researchers sometimes use. Working memory itself has been revised: George Miller's famous "seven plus or minus two" is now often put at about four chunks (Cowan, 2001). DeepMind's Differentiable Neural Computer gave a neural network a separate external memory, which let it learn tasks such as finding routes on the London Underground (Graves et al., 2016). Its authors drew comparisons with biological memory, but it is an engineering system and not a model of human working memory.
Learning and Development: From Child Psychology to Machine Learning
Children can learn a new concept from a few examples, and that gap has motivated a lot of AI research (Lake et al., 2015, 2017). Some projects draw directly on developmental psychology. Piloto and colleagues (2022) adapted the violation-of-expectation method used with infants to build a model that learned intuitive physics, and found that it depended on representing objects. Vong and colleagues (2024) trained a neural network on 61 hours of head-camera recordings from one child, and it learned many word-referent pairings. That shows how much can be learned from a child's own input, without showing that this is how children do it.
Contemporary Convergence: Where AI and Cognitive Science Meet Today
Large Language Models as Cognitive Theories
Large language models (LLMs) can be tested with the same tasks used on people, which has created a new kind of experiment. The results so far are mixed, and they date quickly:
- Binz and Schulz (2023) found that GPT-3 matched or beat people on many classic vignette tasks, but that small changes to the wording could lead it badly astray.
- Kosinski (2024) reported that a 2023 version of ChatGPT solved 75% of a set of false-belief tasks, a test of theory of mind. Ullman (2023) showed that a success of this kind can be reversed by small changes that should not matter to anyone who really tracks beliefs.
- Hagendorff and colleagues (2023) found that earlier GPT models increasingly made human-like intuitive errors as they grew, and that ChatGPT then largely stopped making them.
Passing a test designed for people does not mean a model passes it the way people do. Newer models that write out intermediate reasoning steps do better on many such tasks, so any claim about what LLMs cannot do should carry a date. Researchers remain sharply divided on whether LLMs understand language in any human sense (Mitchell & Krakauer, 2023). Bender and colleagues (2021) argued that fluent text is not evidence of meaning, and others see understanding of a different kind. Frank (2023a) recommends testing such capacities with care, using the same standards developmental psychologists apply to children. The models' training also differs from a child's: they get far more text than a child hears and none of the social and physical experience (Frank, 2023b).
Neuroscience-Inspired Architectures
Some AI work draws on neuroscience more directly. Predictive coding, in which higher levels of a network predict the activity of lower levels (Rao & Ballard, 1999), has inspired alternative learning algorithms and connects to predictive processing theories of the brain. Spiking networks try to capture the timing of real neurons. Whether the brain does anything like backpropagation, the training method behind nearly all modern AI, is an open question (Lillicrap et al., 2020). Mainstream AI systems are not close simulations of the brain, and an artificial "neuron" is a far simpler thing than a biological one.
Shared Challenges and Mutual Solutions
The Common Sense Problem
People rely on a large body of implicit knowledge: objects fall, water flows downhill, other people have beliefs and goals. Psychologists study this directly. Battaglia and colleagues (2013) proposed that people understand physical scenes by running approximate mental simulations. AI researchers have tried to encode such knowledge in resources such as ATOMIC, an atlas of everyday if-then reasoning (Sap et al., 2019). LLMs now answer many common-sense questions correctly, but they can still fail in ways that are hard to predict, and it remains unclear whether they have an underlying model of the world.
Explainability and Interpretability
Understanding how AI systems reach their outputs matters, and research on human explanation can help. Miller (2019) reviewed work from the social sciences and found that people want explanations that are contrastive (why this and not that), selective and social. Tools such as LIME (Ribeiro et al., 2016) highlight which features drove a prediction, and counterfactual explanations say what would have had to change for a different result. These methods offer a simplified account of a model's behavior and should not be mistaken for a full view of how it works.
Ethical Dimensions: Cognitive Science Illuminating AI Ethics
Understanding and Mitigating Bias
AI bias and human cognitive bias are related but different. Human biases such as availability come from shortcuts in thinking. AI bias usually comes from the data. Caliskan and colleagues (2017) showed that word embeddings trained on ordinary web text reproduced a range of human biases measured by the Implicit Association Test, including ones about race and gender. The model absorbs associations already present in the language it learns from. Reducing this takes changes to data, training and how systems are checked.
AI Rights and Moral Status
When Google engineer Blake Lemoine claimed in 2022 that the LaMDA chatbot was sentient, the episode brought an old question to the public: does behaving like a conscious being show that something is conscious? Cognitive science has no agreed test for consciousness, in machines or anywhere else. Theories disagree about whether the right kind of computation would be enough. Anil Seth (2026), for example, argues that consciousness depends on our nature as living organisms and that conscious AI is unlikely along current trajectories. Others are more open to it. Our article on consciousness covers these theories.
Applications and Real-World Impact
Education and Personalized Learning
Intelligent tutoring systems are one of the clearest successes for applied cognitive science. Knowledge tracing, which estimates what a student has mastered, came from cognitive models of skill learning (Corbett & Anderson, 1995). VanLehn (2011) found step-based tutoring systems nearly as effective as human tutors in experiments, and a meta-analysis found a median gain of 0.66 standard deviations, though much smaller results on standardized tests than on locally written ones (Kulik & Fletcher, 2016). At real-world scale, effects are more modest: a large randomized study of Cognitive Tutor Algebra found no effect in the first year and a gain of about eight percentile points in the second (Pane et al., 2014). These systems adapt to what a student knows. They should not be confused with teaching to "learning styles," an idea that lacks good evidence (Pashler et al., 2008).
Language apps apply spaced repetition, and Duolingo has published a model of how it schedules review (Settles & Meeder, 2016). Generative AI results depend heavily on design. In one Harvard physics course, students learned more in less time from a carefully designed AI tutor than from an active-learning class (Kestin et al., 2025). In a field experiment with nearly 1,000 high school students, those given plain GPT-4 access did better during practice but 17% worse once access was removed, while a version with learning safeguards largely avoided that harm (Bastani et al., 2025).
Mental Health and Therapeutic Applications
This area calls for caution. Woebot, a scripted chatbot based on cognitive behavioral therapy, reduced depression symptoms over two weeks in a small trial of 70 students, run by researchers including its founder (Fitzpatrick et al., 2017). The consumer app closed in June 2025, with the founder citing the cost of meeting FDA requirements (Aguilar, 2025). In the first randomized trial of a purpose-built generative chatbot, Therabot, 210 adults showed greater symptom reductions than a waitlist group (Heinz et al., 2025). Clinicians supervised its responses, and the comparison was no treatment, not therapy. In November 2025 the American Psychological Association advised against relying on generative AI chatbots and wellness apps for psychological treatment, saying evidence on effectiveness and safety is insufficient (American Psychological Association, 2025).
Researchers are also studying whether speech patterns can flag depression (Cummins et al., 2015). This remains mostly research, and it is not a diagnostic tool.
Human-AI Collaboration in Complex Domains
Combining people and AI does not automatically give better results. In skin cancer diagnosis, good AI support improved accuracy over either AI or physicians alone, and the least experienced clinicians gained the most, but faulty AI misled clinicians at every level, including experts (Tschandl et al., 2020). A meta-analysis of 106 experiments found that on average human-AI combinations did worse than the better of the two working alone, with losses in decision tasks and gains in content creation (Vaccaro et al., 2024). When to trust a machine is itself a question about human judgment.
Not every AI success involves cognitive science. AlphaFold's protein structure predictions (Jumper et al., 2021) rest on biological data and deep learning, not models of human thought.
The Future: Toward Artificial General Intelligence and Beyond
The Path to AGI
Artificial general intelligence (AGI) usually means systems that match human abilities across most domains. There is no agreed definition or test, and forecasts vary widely. One camp holds that progress has come mostly from general methods and more computation, a view stated in Rich Sutton's essay "The Bitter Lesson" (2019). Another holds that human-like intelligence needs built-in structure such as causal models and intuitive physics (Lake et al., 2017). Cognitive science cannot settle this yet, but it can supply tests that separate real flexibility from good performance on familiar problems.
Augmented Cognition and Brain-Computer Interfaces
Brain-computer interfaces are a working medical technology in early trials, mainly for paralysis. Implanted electrodes have let people with tetraplegia control a robotic arm (Hochberg et al., 2012), and a man with paralysis typed by imagining handwriting (Willett et al., 2021). Neuralink and other companies have begun implanting devices in people with paralysis. Machine learning does the decoding, and these systems read movement intentions. Claims about memory enhancement or direct knowledge transfer are speculation: nothing like them has been demonstrated.
Implications for Human Self-Understanding
As AI takes on tasks once thought to need human intelligence, it raises the question of what is distinctive about human minds. Cognitive science points to features current systems lack or lack evidence for: learning through a body in a social world, caring about outcomes, and acting on goals of one's own. Whether those differences are permanent is an open question, and a live research topic in embodied cognition.
Conclusion: A Two-Way Relationship
AI gives cognitive science models to test and tools to analyze data, and cognitive science gives AI some of its ideas and many of the best methods for evaluating it. The strongest current work treats AI systems as hypotheses to be tested, not as proof of how minds work. If you want to work at this intersection, see our guides to computational cognitive science careers and careers in cognitive science.
Frequently Asked Questions
How is AI related to cognitive science?
Both grew from the 1950s idea that thinking can be studied as information processing. Cognitive scientists use AI models to test theories of the mind, and AI has drawn loose inspiration from psychology and neuroscience. Cognitive science also provides methods for evaluating what AI systems can do.
Does AI think like the human brain?
Not in any established sense. Artificial neural networks were loosely inspired by the brain, and some models predict brain activity well, but they learn differently, need far more data than children and lack a body and social experience. Whether they use similar mechanisms for specific tasks is an active research question.
Do large language models understand language?
Researchers disagree. LLMs pass many tests designed for human understanding, but small changes to a task can make them fail, and passing a test does not show that a model solved it the way a person would. There is no consensus on whether their performance amounts to understanding.
Is AI conscious?
There is no good evidence that current AI systems are conscious, and no agreed test that could settle the question. Researchers disagree about whether computation alone could ever be enough for consciousness. An AI saying that it has feelings is not evidence that it does.
Can AI replace teachers or therapists?
The evidence does not support that. Well-designed tutoring systems can help, but unrestricted chatbot use can harm learning. Some therapy chatbots have shown early benefits in trials, and the American Psychological Association advised in 2025 against relying on generative AI chatbots and wellness apps for psychological treatment.
Sources
- Aguilar, M. (2025, July 2). Why Woebot, a pioneering therapy chatbot, shut down. STAT. statnews.com
- American Psychological Association. (2025). Health advisory: Use of generative AI chatbots and wellness applications for mental health. apa.org
- Bahdanau, D., Cho, K., & Bengio, Y. (2014). Neural machine translation by jointly learning to align and translate. arXiv. arXiv:1409.0473
- Bastani, H., Bastani, O., Sungu, A., Ge, H., Kabakcı, Ö., & Mariman, R. (2025). Generative AI without guardrails can harm learning: Evidence from high school mathematics. Proceedings of the National Academy of Sciences, 122(26), e2422633122. doi:10.1073/pnas.2422633122
- Battaglia, P. W., Hamrick, J. B., & Tenenbaum, J. B. (2013). Simulation as an engine of physical scene understanding. Proceedings of the National Academy of Sciences, 110(45), 18327–18332. doi:10.1073/pnas.1306572110
- Bender, E. M., Gebru, T., McMillan-Major, A., & Shmitchell, S. (2021). On the dangers of stochastic parrots: Can language models be too big? Proceedings of the 2021 ACM Conference on Fairness, Accountability, and Transparency, 610–623. doi:10.1145/3442188.3445922
- Binz, M., & Schulz, E. (2023). Using cognitive psychology to understand GPT-3. Proceedings of the National Academy of Sciences, 120(6), e2218523120. doi:10.1073/pnas.2218523120
- Caliskan, A., Bryson, J. J., & Narayanan, A. (2017). Semantics derived automatically from language corpora contain human-like biases. Science, 356(6334), 183–186. doi:10.1126/science.aal4230
- Corbett, A. T., & Anderson, J. R. (1995). Knowledge tracing: Modeling the acquisition of procedural knowledge. User Modelling and User-Adapted Interaction, 4(4), 253–278. doi:10.1007/BF01099821
- Cowan, N. (2001). The magical number 4 in short-term memory: A reconsideration of mental storage capacity. Behavioral and Brain Sciences, 24(1), 87–114. doi:10.1017/S0140525X01003922
- Cummins, N., Scherer, S., Krajewski, J., Schnieder, S., Epps, J., & Quatieri, T. F. (2015). A review of depression and suicide risk assessment using speech analysis. Speech Communication, 71, 10–49. doi:10.1016/j.specom.2015.03.004
- Fitzpatrick, K. K., Darcy, A., & Vierhile, M. (2017). Delivering cognitive behavior therapy to young adults with symptoms of depression and anxiety using a fully automated conversational agent (Woebot): A randomized controlled trial. JMIR Mental Health, 4(2), e19. doi:10.2196/mental.7785
- Frank, M. C. (2023a). Baby steps in evaluating the capacities of large language models. Nature Reviews Psychology, 2(8), 451–452. doi:10.1038/s44159-023-00211-x
- Frank, M. C. (2023b). Bridging the data gap between children and large language models. Trends in Cognitive Sciences, 27(11), 990–992. doi:10.1016/j.tics.2023.08.007
- Fukushima, K. (1980). Neocognitron: A self-organizing neural network model for a mechanism of pattern recognition unaffected by shift in position. Biological Cybernetics, 36(4), 193–202. doi:10.1007/BF00344251
- Graves, A., Wayne, G., Reynolds, M., Harley, T., Danihelka, I., Grabska-Barwińska, A., et al. (2016). Hybrid computing using a neural network with dynamic external memory. Nature, 538(7626), 471–476. doi:10.1038/nature20101
- Hagendorff, T., Fabi, S., & Kosinski, M. (2023). Human-like intuitive behavior and reasoning biases emerged in large language models but disappeared in ChatGPT. Nature Computational Science, 3(10), 833–838. doi:10.1038/s43588-023-00527-x
- Hassabis, D., Kumaran, D., Summerfield, C., & Botvinick, M. (2017). Neuroscience-inspired artificial intelligence. Neuron, 95(2), 245–258. doi:10.1016/j.neuron.2017.06.011
- Heinz, M. V., Mackin, D. M., Trudeau, B. M., Bhattacharya, S., Wang, Y., Banta, H. A., et al. (2025). Randomized trial of a generative AI chatbot for mental health treatment. NEJM AI, 2(4). doi:10.1056/AIoa2400802
- Hochberg, L. R., Bacher, D., Jarosiewicz, B., Masse, N. Y., Simeral, J. D., Vogel, J., et al. (2012). Reach and grasp by people with tetraplegia using a neurally controlled robotic arm. Nature, 485(7398), 372–375. doi:10.1038/nature11076
- Jumper, J., Evans, R., Pritzel, A., Green, T., Figurnov, M., Ronneberger, O., et al. (2021). Highly accurate protein structure prediction with AlphaFold. Nature, 596(7873), 583–589. doi:10.1038/s41586-021-03819-2
- Kestin, G., Miller, K., Klales, A., Milbourne, T., & Ponti, G. (2025). AI tutoring outperforms in-class active learning: an RCT introducing a novel research-based design in an authentic educational setting. Scientific Reports, 15(1), 17458. doi:10.1038/s41598-025-97652-6
- Kosinski, M. (2024). Evaluating large language models in theory of mind tasks. Proceedings of the National Academy of Sciences, 121(45), e2405460121. doi:10.1073/pnas.2405460121
- Kulik, J. A., & Fletcher, J. D. (2016). Effectiveness of intelligent tutoring systems: A meta-analytic review. Review of Educational Research, 86(1), 42–78. doi:10.3102/0034654315581420
- Lake, B. M., Salakhutdinov, R., & Tenenbaum, J. B. (2015). Human-level concept learning through probabilistic program induction. Science, 350(6266), 1332–1338. doi:10.1126/science.aab3050
- Lake, B. M., Ullman, T. D., Tenenbaum, J. B., & Gershman, S. J. (2017). Building machines that learn and think like people. Behavioral and Brain Sciences, 40, e253. doi:10.1017/S0140525X16001837
- Lieder, F., & Griffiths, T. L. (2020). Resource-rational analysis: Understanding human cognition as the optimal use of limited computational resources. Behavioral and Brain Sciences, 43, e1. doi:10.1017/S0140525X1900061X
- Lillicrap, T. P., Santoro, A., Marris, L., Akerman, C. J., & Hinton, G. (2020). Backpropagation and the brain. Nature Reviews Neuroscience, 21(6), 335–346. doi:10.1038/s41583-020-0277-3
- Lindsay, G. W. (2020). Attention in psychology, neuroscience, and machine learning. Frontiers in Computational Neuroscience, 14, 29. doi:10.3389/fncom.2020.00029
- Long, B., Fan, J. E., Huey, H., Chai, Z., & Frank, M. C. (2024). Parallel developmental changes in children’s production and recognition of line drawings of visual concepts. Nature Communications, 15(1), 1191. doi:10.1038/s41467-023-44529-9
- Marr, D. (1982). Vision. W. H. Freeman.
- Miller, T. (2019). Explanation in artificial intelligence: Insights from the social sciences. Artificial Intelligence, 267, 1–38. doi:10.1016/j.artint.2018.07.007
- Mitchell, M., & Krakauer, D. C. (2023). The debate over understanding in AI’s large language models. Proceedings of the National Academy of Sciences, 120(13), e2215907120. doi:10.1073/pnas.2215907120
- Newell, A., Shaw, J. C., & Simon, H. A. (1958). Elements of a theory of human problem solving. Psychological Review, 65(3), 151–166. doi:10.1037/h0048495
- Pane, J. F., Griffin, B. A., McCaffrey, D. F., & Karam, R. (2014). Effectiveness of Cognitive Tutor Algebra I at scale. Educational Evaluation and Policy Analysis, 36(2), 127–144. doi:10.3102/0162373713507480
- Pashler, H., McDaniel, M., Rohrer, D., & Bjork, R. (2008). Learning styles. Psychological Science in the Public Interest, 9(3), 105–119. doi:10.1111/j.1539-6053.2009.01038.x
- Piloto, L. S., Weinstein, A., Battaglia, P., & Botvinick, M. (2022). Intuitive physics learning in a deep-learning model inspired by developmental psychology. Nature Human Behaviour, 6(9), 1257–1267. doi:10.1038/s41562-022-01394-8
- Rao, R. P. N., & Ballard, D. H. (1999). Predictive coding in the visual cortex: a functional interpretation of some extra-classical receptive-field effects. Nature Neuroscience, 2(1), 79–87. doi:10.1038/4580
- Ribeiro, M. T., Singh, S., & Guestrin, C. (2016). “Why should I trust you?”: Explaining the predictions of any classifier. Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, 1135–1144. doi:10.1145/2939672.2939778
- Sap, M., Le Bras, R., Allaway, E., Bhagavatula, C., Lourie, N., Rashkin, H., et al. (2019). ATOMIC: An atlas of machine commonsense for if-then reasoning. Proceedings of the AAAI Conference on Artificial Intelligence, 33(01), 3027–3035. doi:10.1609/aaai.v33i01.33013027
- Schrimpf, M., Blank, I. A., Tuckute, G., Kauf, C., Hosseini, E. A., Kanwisher, N., et al. (2021). The neural architecture of language: Integrative modeling converges on predictive processing. Proceedings of the National Academy of Sciences, 118(45), e2105646118. doi:10.1073/pnas.2105646118
- Seth, A. K. (2026). Conscious artificial intelligence and biological naturalism. Behavioral and Brain Sciences, 49, e315. doi:10.1017/S0140525X25000032
- Settles, B., & Meeder, B. (2016). A trainable spaced repetition model for language learning. Proceedings of the 54th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), 1848–1858. doi:10.18653/v1/p16-1174
- Sutton, R. (2019). The bitter lesson. incompleteideas.net
- Treisman, A. M., & Gelade, G. (1980). A feature-integration theory of attention. Cognitive Psychology, 12(1), 97–136. doi:10.1016/0010-0285(80)90005-5
- Tschandl, P., Rinner, C., Apalla, Z., Argenziano, G., Codella, N., Halpern, A., et al. (2020). Human–computer collaboration for skin cancer recognition. Nature Medicine, 26(8), 1229–1234. doi:10.1038/s41591-020-0942-0
- Turing, A. M. (1950). Computing machinery and intelligence. Mind, 59(236), 433–460. doi:10.1093/mind/LIX.236.433
- Ullman, T. (2023). Large language models fail on trivial alterations to theory-of-mind tasks. arXiv. arXiv:2302.08399
- Vaccaro, M., Almaatouq, A., & Malone, T. (2024). When combinations of humans and AI are useful: A systematic review and meta-analysis. Nature Human Behaviour, 8(12), 2293–2303. doi:10.1038/s41562-024-02024-1
- VanLehn, K. (2011). The relative effectiveness of human tutoring, intelligent tutoring systems, and other tutoring systems. Educational Psychologist, 46(4), 197–221. doi:10.1080/00461520.2011.611369
- Vaswani, A., Shazeer, N., Parmar, N., Uszkoreit, J., Jones, L., Gomez, A. N., Kaiser, L., & Polosukhin, I. (2017). Attention is all you need. arXiv. arXiv:1706.03762
- Vong, W. K., Wang, W., Orhan, A. E., & Lake, B. M. (2024). Grounded language acquisition through the eyes and ears of a single child. Science, 383(6682), 504–511. doi:10.1126/science.adi1374
- Willett, F. R., Avansino, D. T., Hochberg, L. R., Henderson, J. M., & Shenoy, K. V. (2021). High-performance brain-to-text communication via handwriting. Nature, 593(7858), 249–254. doi:10.1038/s41586-021-03506-2
- Yamins, D. L. K., Hong, H., Cadieu, C. F., Solomon, E. A., Seibert, D., & DiCarlo, J. J. (2014). Performance-optimized hierarchical models predict neural responses in higher visual cortex. Proceedings of the National Academy of Sciences, 111(23), 8619–8624. doi:10.1073/pnas.1403112111