Lecturer: Marianna Bolognesi
Fields: Cognitive Science, Linguistics, AI
Content
This course focuses on the (human) ability to abstract from experience and from language, to construct higher-order representations that are used to reason, form judgments, and appreciate art among other things. We focus on the critical role of language in performing abstractions and generalizations. Students will learn about the distinction between concreteness and specificity in word processing, uncovering how these variables shape linguistic contexts and influence thought. The course will address groundbreaking research that challenges traditional views, highlighting the overlooked role of specificity in language variability. Additionally, students will analyze the limitations of Large Language Models (LLMs), particularly their inability to accurately interpret generic statements and their tendency to overgeneralize, potentially reinforcing stereotypes. Through interdisciplinary discussions spanning cognitive science, psycholinguistics, and AI ethics, the course will provide students with a comprehensive understanding of abstraction’s significance and its implications for advancing human-like AI systems.
Literature
- Barsalou LW. (2003). Abstraction in perceptual symbol systems. Philos Trans R Soc Lond B Biol Sci. 29;358(1435):1177-87.
- Bolognesi, M., Burgers, C., & Caselli, T. (2020). On abstraction: Decoupling conceptual concreteness and categorical specificity. Cognitive Processing, 21(3), 365–381.
- Collacciani, C., Rambelli, G. and Bolognesi, M. (2024). Quantifying Generalizations: Exploring the Divide Between Human and LLMs’ Sensitivity to Quantification. In Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 11811–11822, Bangkok, Thailand. Association for Computational Linguistics.
- Rambelli, G. & Bolognesi, M. (2024). The Contextual Variability of English Nouns: The Impact of Categorical Specificity beyond Conceptual Concreteness. In Proceedings of the 2024 Joint International Conference on Computational Linguistics, Language Resources and Evaluation (LREC-COLING 2024). Pp: 15854–15860.
- Rissman, L., & Lupyan, G. (2024). Words do not just label concepts: activating superordinate categories through labels, lists, and definitions. Language, Cognition and Neuroscience, 39(5), 657–676.
Lecturer
Linguist (cognitive and distributional semantics). She was a Marie S. Curie research fellow at the University of Amsterdam (2015-2017), research associate at the University of Oxford (2017-2019), and now associate professor at the University of Bologna, Italy. In 2022 she won an ERC grant for the project ABSTRACTION ERC-2021-STG-101039777,which aims to investigate how abstraction mechanisms work in thought, verbal language and creativity, both human and generated by artificial intelligence. She is vice-PI and work unit coordinator of the national project PRIN 2022 “WEMB: Word Embeddings from Cognitive Linguistics to Language Engineering and Back”, a project that aims to understand how vector representations of word meaning (embeddings) reflect those in the minds of the speakers. Her research employs cross-disciplinary approaches, combining psycholinguistic experiments and computational modeling.
Affiliation: University of Bologna, Italy
Homepage: https://www.unibo.it/sitoweb/m.bolognesi/en