Call for Papers: Special Issue on Commonsense Reasoning
Submitted by Pascal Hitzler on
Common sense can be defined as the broad set of knowledge and inferential skills shared between people, who seamlessly use it in understanding the everyday world of objects and their interactions, at the physical, emotional, and social levels. When we look at mainstream AI, such capability is still fundamentally missing: for instance, Large Language Models have recently become popular by demonstrating remarkable fluency in conversing with humans, but they still make trivial mistakes when probed for commonsense competence. This problem hinders progress on trustworthy AI that is robust, explainable, and collaborative. Meanwhile, the AI community is turning attention towards hybrid systems that combine neural and symbolic methods, or NeSy. By bringing together these two approaches, the promise of NeSy is to improve the robustness and explainability of AI agents, and one of the key aspects of intelligence required for this is exactly common sense. NeSy lends itself as a framework that can also connect empirical modeling pursuits with background cognitive principles, theories, and architectures, and leverage its symbolic outlet to make an impact across application domains and collaborate with people in a responsible and sustainable manner.
The goal of this Special Issue is to bring together a wide set of ideas on how to develop, evaluate, and apply AI methods and agents endowed with common sense. This pursuit is inherently multidimensional and interdisciplinary. Commonsense methods and agents require technical innovations on NeSy architectures that may include LLMs, but also go beyond a strictly-linguistic focus. A long list of ideas from cognitive science are of interest too, including prototype theory, mental models, analogical reasoning, and reasoning under bounded rationality. Reliable evaluation requires novel commonsense tasks, which are as realistic as possible (e.g., through being interactive in nature), but also out-of-the-box ideas for evaluating commonsense abilities, such as novel metrics and simulation environments. More research is needed to understand or improve how properties like explainability, robustness, collaboration, and responsibility apply to machine common sense. Commonsense reasoning has recently inspired various new trends such as Anticipatory Thinking, Meta-Cognition, Causality and Counterfactual Reasoning, Harnessing Commonsense from Different Modalities, Multilinguality, Lateral Thinking, and Human-AI Teaming. Finally, a key aspect of common sense is its real-world application - while much of the present research focuses on general, open-domain reasoning, many closed-domains including autonomous driving, healthcare, and social media, would benefit significantly from deep reasoning systems that incorporate common sense.
Deadline
November 30th, 2024. Earlier submissions will be processed as they come in.
Guest editors
Filip Ilievski (USC Information Sciences Institute / Vrije Universiteit Amsterdam)
Alessandro Oltramari (Bosch Center for Artificial Intelligence)
Knowledge-infused LLMs for Natural Language Understanding
New architectures for NeSy commonsense reasoning
Cognitively-inspired Commonsense Reasoning
Reasoning under bounded rationality
Reasoning with prototypes
Analogical reasoning
Reasoning under uncertainty
Integration between cognitive architecture and neuro-symbolic methods
Interactive commonsense tasks
Conversational Systems
Q/A tasks
Text-based Games
Story understanding and generation
Evaluating commonsense reasoning
Systematic probing
Simulation environments
Novel tasks and metrics
Commonsense properties of AI systems
Explainability/Interpretability
Robustness/Adaptivity
Collaborative AI
Responsible AI
New Trends
Anticipatory Thinking
Causality and Counterfactual Reasoning
Harnessing Commonsense from Different Modalities
Multilinguality
Lateral Thinking
Human-AI Teaming
Applications of commonsense reasoning
Guest Editorial Board
(to be completed)
Author Guidelines
We invite full papers, dataset descriptions, application reports and reports on tools and systems. Submissions must be original and should not have been published previously or be under consideration for publication while being evaluated for this special issue. Authors can extend previously published conference or workshop papers - see the submission guidelines at https://neurosymbolic-ai-journal.com/content/author-guidelines for details.
Note that you need to request an account on the website for submitting a paper. Please indicate in the cover letter that it is for the "Special Issue on Commonsense Reasoning". All manuscripts will be reviewed based on the journal's open and transparent review policy and will be made available online during the review process.
Common sense can be defined as the broad set of knowledge and inferential skills shared between people, who seamlessly use it in understanding the everyday world of objects and their interactions, at the physical, emotional, and social levels. When we look at mainstream AI, such capability is still fundamentally missing: for instance, Large Language Models have recently become popular by demonstrating remarkable fluency in conversing with humans, but they still make trivial mistakes when probed for commonsense competence. This problem hinders progress on trustworthy AI that is robust, explainable, and collaborative. Meanwhile, the AI community is turning attention towards hybrid systems that combine neural and symbolic methods, or NeSy. By bringing together these two approaches, the promise of NeSy is to improve the robustness and explainability of AI agents, and one of the key aspects of intelligence required for this is exactly common sense. NeSy lends itself as a framework that can also connect empirical modeling pursuits with background cognitive principles, theories, and architectures, and leverage its symbolic outlet to make an impact across application domains and collaborate with people in a responsible and sustainable manner.
Deadline
November 30th, 2024. Earlier submissions will be processed as they come in.
Guest editors
Topics of Interest include
Guest Editorial Board
Author Guidelines