{"slug": "ai-framework-rooted-in-cognitive-science-could-complete-tasks-more-efficiently", "title": "AI framework rooted in cognitive science could complete tasks more efficiently", "summary": "Researchers at Tsinghua University, Graz University of Technology, and the National Research Council in Italy have developed an AI framework inspired by cognitive science that uses cognitive maps to enable more efficient problem-solving, potentially reducing the energy consumption of AI systems. The framework, published in Nature Machine Intelligence, was used to create an artificial neural network that can be deployed on energy-efficient neuromorphic devices, addressing the high computational and electricity demands of current deep neural networks and large language models.", "body_md": "August 8, 2026\n[\nfeature\n](https://techxplore.com/editorials/)\n\n# AI framework rooted in cognitive science could complete tasks more efficiently\n\n##### Ingrid Fadelli\n\nAuthor\n\n##### Robert Egan\n\nSenior Editor\n\nIn recent years, computer scientists have developed a wide range of artificial intelligence (AI) models that can rapidly recognize patterns in data, generate content and solve other computational problems. Many of these AI systems are based on deep neural networks (DNNs), brain-inspired computational models that can make predictions based on specific data, or LLMs, models that can process human language, answer queries and generate text.\n\nWhile both DNNs and LLMs often perform remarkably well, they generally require vast computational resources and consume large amounts of electricity. Some research teams have thus been trying to develop new models or computational strategies that could reduce the energy consumption associated with AI.\n\nResearchers at Tsinghua University, Graz University of Technology and the National Research Council in Italy introduced a new AI framework that draws inspiration from ideas rooted in cognitive science and neuroscience. Their proposed approach, introduced in a paper [published](https://www.nature.com/articles/s42256-026-01254-4) in *Nature Machine Intelligence*, was used to develop an artificial neural network that could solve problems more adaptively and could be easily deployed on energy-efficient neuromorphic devices.\n\n\"We were intrigued by the fact that evolution had invented algorithms and data structures that produce intelligence in brains, and that these solutions that nature had found differ strongly from those that are used to produce intelligence in current AI,\" Wolfgang Maass, senior author of the paper, told Tech Xplore.\n\n\"What do they look like? Can we reproduce them in artificial devices? The latter is of particular importance because the energy consumption of current AI tools has become a major economic, political and scientific impediment, whereas the brain needs only 20W for producing intelligence.\"\n\n## Creating AI inspired by the brain's problem-solving strategies\n\nCognitive scientists and neuroscientists recently collected various precise recordings of neural activity in the human brain while people solved different types of problems. These brain recordings and the insights derived from them could be valuable for developing brain-inspired computational models and algorithms.\n\n\"It is very tempting to use these hints for designing corresponding algorithms in artificial systems,\" Maass explained. \"We were especially interested in reconstructing algorithmic approaches of the brain for solving difficult problems, which require some form of intuition or fantasy. We wondered whether we could find ways to reproduce this in a simple chip, without using DNNs or LLMs.\"\n\nOngoing cognitive science research suggests that the brain tends to [encode learned knowledge](https://medicalxpress.com/news/2023-04-brains-decisions-virtual-monster-world.html?utm_source=embeddings&utm_medium=related&utm_campaign=internal) in the form of representations known as cognitive maps, as opposed to tables or series of parameters. These cognitive maps help the brain respond both rapidly and adaptively while trying to solve problems in real time.\n\nInterestingly, most AI systems developed to date do not encode information in this fashion. Maass and his colleagues wanted to fill this gap in the literature by developing a new AI framework that leverages cognitive maps to organize information.\n\n\"There already exists a very large number of studies that investigate brain algorithms at the micro level by considering computational properties of biological neurons,\" Maass said. \"Our results show that experimental data from the brain also provide novel insight at a higher level of computational organization and data structures.\n\n\"In contrast to current AI methods, these brain-like methods can easily be implemented in [novel types of chips](https://techxplore.com/news/2025-07-ai-energy-efficient-human-brain.html?utm_source=embeddings&utm_medium=related&utm_campaign=internal) that aim to make AI implementations more energy efficient. These are currently being developed under the names in-memory computing chips and neuromorphic chips by several large companies (IBM, Intel) and numerous startups.\"\n\nThe researchers have already carried out initial tests to assess their proposed approach. They found that a neural network based on their approach could plan adaptively and successfully solve problems it had never encountered before. Moreover, the processes guiding its predictions and actions were easy to interpret compared with those of many existing AI models.\n\n## Next steps in the development of brain-like AI\n\nThe new AI model introduced by this research team could soon be further refined and tested on a wider range of problems. A key advantage of the team's model is that it learns locally and does not require computationally intensive training processes.\n\nIn the future, the model could be deployed on neuromorphic computing chips and in-memory computing systems. [In-memory computing systems](https://techxplore.com/news/2022-08-synapses-solid-state-memory-neuromorphic-circuits.html?utm_source=embeddings&utm_medium=related&utm_campaign=internal) avoid shuffling data between memory components and processors, which tends to consume significant energy, by processing data directly in the storage system (e.g., in a memristor array).\n\nNotably, Maass and his colleagues are now working with engineers at Intel and another startup in the U.S. to implement their brain-like algorithm on chips.\n\n\"We are continuing our work on porting brain-like algorithms into more energy-efficient hardware,\" Maass added. \"In our forthcoming work, we will show that we can also capture brain-like low latency and our astounding flexibility of decision-making in the face of new goals or contingencies. We are also able to capture that our brains usually automatically provide explanations for chosen actions in the form of concrete experiences that support a decision. Notably, making AI decisions explainable is a major goal in current AI research.\"\n\nWritten for you by our author [Ingrid Fadelli](https://sciencex.com/help/editorial-team/ingrid-fadelli/), edited by [Robert Egan](https://sciencex.com/help/editorial-team/robert-egan/)—this article is the result of careful human work. We rely on readers like you to keep independent science journalism alive.\nIf this reporting matters to you, please consider a [donation](https://sciencex.com/donate/?utm_source=story&utm_medium=story&utm_campaign=story) (especially monthly). You'll get an **ad-free** account as a thank-you.\n\n###### Publication details\n\nHui Lin et al, Neural sampling from cognitive maps enables goal-directed imagination and planning, *Nature Machine Intelligence* (2026). [DOI: 10.1038/s42256-026-01254-4](https://dx.doi.org/10.1038/s42256-026-01254-4).\n\n**Journal information:**\n[Nature Machine Intelligence](https://techxplore.com/journals/nature-machine-intelligence/)\n[\n](https://www.nature.com/natmachintell/)\n\n© 2026 Science X Network\n\n**Citation**: AI framework rooted in cognitive science could complete tasks more efficiently (2026, August 8) retrieved 10 August 2026 from https://techxplore.com/news/2026-07-ai-framework-rooted-cognitive-science.html", "url": "https://wpnews.pro/news/ai-framework-rooted-in-cognitive-science-could-complete-tasks-more-efficiently", "canonical_source": "https://techxplore.com/news/2026-07-ai-framework-rooted-cognitive-science.html", "published_at": "2026-08-10 22:04:39+00:00", "updated_at": "2026-08-10 22:41:33.692121+00:00", "lang": "en", "topics": ["artificial-intelligence", "neural-networks", "ai-research"], "entities": ["Tsinghua University", "Graz University of Technology", "National Research Council in Italy", "Nature Machine Intelligence", "Wolfgang Maass"], "alternates": {"html": "https://wpnews.pro/news/ai-framework-rooted-in-cognitive-science-could-complete-tasks-more-efficiently", "markdown": "https://wpnews.pro/news/ai-framework-rooted-in-cognitive-science-could-complete-tasks-more-efficiently.md", "text": "https://wpnews.pro/news/ai-framework-rooted-in-cognitive-science-could-complete-tasks-more-efficiently.txt", "jsonld": "https://wpnews.pro/news/ai-framework-rooted-in-cognitive-science-could-complete-tasks-more-efficiently.jsonld"}}