Thinking Fast and Slow in AI: The Role of Metacognition

Ganapini, Marianna Bergamaschi; Campbell, Murray; Fabiano, Francesco; Horesh, Lior; Lenchner, Jon; Loreggia, Andrea; Mattei, Nicholas; Rossi, Francesca; Srivastava, Biplav; Venable, Kristen Brent

doi:10.1007/978-3-031-25891-6_38

Cited by 8 publications

(3 citation statements)

References 24 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…A recent example is cognitive emulation, which aims to build a more understandable and controllable model that follows a more human way of reasoning and handling failure. Current efforts include a model architecture based on the “thinking fast and slow theory” to constrain AI systems to make decisions in a constrained, but more human‐like and understandable environment 80 …”

Section: Discussionmentioning

confidence: 99%

Potential merits and flaws of large language models in epilepsy care: A critical review

van Diessen,

van Amerongen,

Zijlmans

et al. 2024

Epilepsia

View full text Add to dashboard Cite

The current pace of development and applications of large language models (LLMs) is unprecedented and will impact future medical care significantly. In this critical review, we provide the background to better understand these novel artificial intelligence (AI) models and how LLMs can be of future use in the daily care of people with epilepsy. Considering the importance of clinical history taking in diagnosing and monitoring epilepsy—combined with the established use of electronic health records—a great potential exists to integrate LLMs in epilepsy care. We present the current available LLM studies in epilepsy. Furthermore, we highlight and compare the most commonly used LLMs and elaborate on how these models can be applied in epilepsy. We further discuss important drawbacks and risks of LLMs, and we provide recommendations for overcoming these limitations.

show abstract

Section: Discussionmentioning

confidence: 99%

Potential merits and flaws of large language models in epilepsy care: A critical review

van Diessen,

van Amerongen,

Zijlmans

et al. 2024

Epilepsia

View full text Add to dashboard Cite

show abstract

“…For these reasons, a growing segment of the AI community is attempting to address these limitations and is trying to create systems that display more "human-like qualities". One of the central strategies to tackle this problem, adopted by various research groups Newell [1992], Goel et al [2017], Ganapini et al [2021], envisions tools, usually referred to as cognitive architectures Kotseruba and Tsotsos [2020], that exploit a combination of both the aforementioned approaches. In particular, in this paper, we explore classical and multi-agent epistemic planning in the context of one of these architectures that stems from a modern cognitive theory, i.e., the well-known Thinking Fast and Slow paradigm proposed by D. Kahneman Kahneman [2011].…”

Section: Thinking Fast and Slowmentioning

confidence: 99%

Fast and Slow Planning

Fabiano¹,

Pallagani²,

Ganapini³

et al. 2023

Preprint

View full text Add to dashboard Cite

The concept of Artificial Intelligence has gained a lot of attention over the last decade. In particular, AI-based tools have been employed in several scenarios and are, by now, pervading our everyday life. Nonetheless, most of these systems lack many capabilities that we would naturally consider to be included in a notion of "intelligence". In this work, we present an architecture that, inspired by the cognitive theory known as Thinking Fast and Slow by D. Kahneman, is tasked with solving planning problems in different settings, specifically: classical and multi-agent epistemic. The system proposed is an instance of a more general AI paradigm, referred to as SOFAI (for Slow and Fast AI). SOFAI exploits multiple solving approaches, with different capabilities that characterize them as either fast or slow, and a metacognitive module to regulate them. This combination of components, which roughly reflects the human reasoning process according to D. Kahneman, allowed us to enhance the reasoning process that, in this case, is concerned with planning in two different settings. The behavior of this system is then compared to state-of-the-art solvers, showing that the newly introduced system presents better results in terms of generality, solving a wider set of problems with an acceptable trade-off between solving times and solution accuracy.

show abstract

“…Anthony et al (2017) apply them to augment reinforcement learning algorithms with deliberative planning of policies through tree search. Ganapini et al (2022) combine them for more efficient navigation with AI agents that evolve from slow to fast decision-making while navigating. Similarly inspired by dual process theories, Hua and Zhang (2022) apply logical reasoning over representation learning for more accurate commonsense knowledge base completion.…”

Section: Related Workmentioning

confidence: 99%

From Heuristic to Analytic: Cognitively Motivated Strategies for Coherent Physical Commonsense Reasoning

Zhang,

Storks,

et al. 2023

Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing

View full text Add to dashboard Cite

Pre-trained language models (PLMs) have shown impressive performance in various language tasks. However, they are prone to spurious correlations, and often generate illusory information. In real-world applications, PLMs should justify decisions with formalized, coherent reasoning chains, but this challenge remains under-explored. Cognitive psychology theorizes that humans are capable of utilizing fast and intuitive heuristic thinking to make decisions based on past experience, then rationalizing the decisions through slower and deliberative analytic reasoning. We incorporate these interlinked dual processes in fine-tuning and in-context learning with PLMs, applying them to two language understanding tasks that require coherent physical commonsense reasoning. We show that our proposed Heuristic-Analytic Reasoning (HAR) strategies drastically improve the coherence of rationalizations for model decisions, yielding state-of-the-art results on Tiered Reasoning for Intuitive Physics (TRIP). We also find that this improved coherence is a direct result of more faithful attention to relevant language context in each step of reasoning. Our findings suggest that human-like reasoning strategies can effectively improve the coherence and reliability of PLM reasoning. Story A: 1. Air is brought in through the mouth. 2. Passes through the lungs. 3. And into the bronchial tissue. 4. The carbon dioxide is removed. 5. The lungs bring the oxygen to the rest of the body. Story B: 1. Carbon dioxide enters the leaves through the stomates by diffusion. 2. Water is transported to the leaves in the xylem. 3. Energy harvested through light reaction is stored by forming ATP. 4. Carbon dioxide and energy from ATP are used to create sugar. 5. Oxygen exits the leaves through the stomata by diffusion. … Story A: 1. Mary went to the fridge. 2. Mary took out a bowl from the fridge. 3. The bowl had a cucumber and a donut in it. 4. Mary put the cucumber on the counter. 5. Mary ate the donut. Story B: 1. Mary went to the fridge. 2. Mary took out a bowl from the fridge. 3. The bowl had a cucumber and a donut in it. 4. Mary tossed the donut in the trash. 5. Mary ate the donut. Plausible story: A Conflicting sentences: (4, 5) States: inedible(donut) → edible(donut) Carbon dioxide conversion story: B Carbon dioxide conversion sentence: 4 Carbon dioxide conversion entity: sugarStory A: 41.0% 1. Coal is heated in the boiler. 2. The water tank over the boiler is heated. 3. Creates steam. 4. The steam is funneled to the piston. 5. Piston uses the steam as energy. 6. The piston causes the crankshaft to move. Story B: 59.0% 1. Plates on the Earth's crust move slowly past each other. 2. As the plates move, they exert a great force. 3. When the force is large enough, the crust breaks. 4. The stress is released as energy. 5. The energy moves through the Earth in the form of waves. 6. We feel the earthquake. What happened to energy? Story A: 16.3% 1. Coal is heated in the boiler. 2. The water tank over the boiler is heated. 3. Creates steam. 4. The steam is funne...

show abstract

Thinking Fast and Slow in AI: The Role of Metacognition

Cited by 8 publications

References 24 publications

Potential merits and flaws of large language models in epilepsy care: A critical review

Potential merits and flaws of large language models in epilepsy care: A critical review

Fast and Slow Planning

From Heuristic to Analytic: Cognitively Motivated Strategies for Coherent Physical Commonsense Reasoning

Contact Info

Product

Resources

About