To promote interdisciplinary interaction between different faculty members and students on the campus, a weekly meeting has been organized by our postdoc Jialiang Yan since September 2023, called Tsing Ch’a Sessions (清茶会). Its slogan “know thyself and let others know you better.”
■Current Sessions
2026 May 28 16:00-17:30 Weijun Yu (余伟俊, Tsinghua University) The Logic of Shuo Zai in the Mohist Canons
The Mohists attached great importance to shuo 說 (“explanation”) and proposed the principle of “using explanations to bring out reasons” (以說出故). In the Mohist Canons, especially the Jingxia (Canon B), the sentence structure shuo zai 說在 + X appears frequently and may be regarded as a linguistic manifestation of this principle.
This talk begins with a brief comparison between shuo (“explanation”) and a group of related concepts—argument, justification, and proof—and then examines the function and grammar of shuo and shuo zai in Mohism. The presentation focuses on a systematic classification of the semantic types of the shuo zai structure and examines how it relates to core Mohist concepts such as lei 類(“kind”), gu 故(“reason/cause”), and li 理(“principle”).
2026 May 14 14:00-15:30 Yumin Ji (池幽旻, Tsinghua University) Language, Logic, and Cognition: Generics, Medvedev Logic, and Learning via limited instances
Generics are sentences that express generalizations while tolerating exceptions, as in “Birds fly” or “Sharks attack people.” Despite their ubiquity in everyday language, the cognitive mechanisms underlying their use remain unclear. This question has attracted attention across several disciplines, particularly linguistics, logic, and cognitive science. This talk is situated at the intersection of these three fields.
The talk begins with a brief overview of the historical development of research on generics in linguistics, followed by a cross-linguistic analysis identifying three central notions associated with generics: typicality, diagnosticity, and naturalness. I then introduce Gärdenfors‘s theory of conceptual spaces [Gärdenfors 2000, 2014], which aims to model how humans learn, structure, and organize concepts. Building on this framework, I propose a modification that allows a more flexible representation of concept formation and aligns with an instance-based view of learning. The three key notions associated to generics are then formally represented within the revised framework. I also present an intriguing relationship between the modified framework and Medvedev logic, an intermediate logic that is not finitely axiomatizable. Finally, I examine whether the proposed formalization aligns with empirical findings from cognitive science.
References
[Gärdenfors 2000] Gärdenfors, P. (2000). Conceptual Spaces: The Geometry of Thought. MIT Press, Cambridge, MA.
[Gärdenfors 2014] Gärdenfors, P. (2014). The Geometry of Meaning: Semantics Based on Conceptual Spaces. MIT Press, Cambridge, MA.
2026 Apr 9 14:00-15:30 Wenlong Zheng (郑文龙, Tsinghua University) Nonmonotonicity and Actual Causation: The Theory of Bochman’s Causal Calculus
Causality is a fundamental concept in reasoning, yet capturing its exact formal semantics has historically posed a challenge for standard deductive systems. In his 2021 book A Logical Theory of Causality, Alexander Bochman develops a causal calculus based on rules of the form A\Rightarrow B (“A causes B”), situated within a nonmonotonic reasoning framework. By treating causal rules as default assumptions, we will see how causal reasoning emerges as a special instance of nonmonotonic reasoning developed within the knowledge representation and symbolic branches of artificial intelligence.
In this talk, we will first introduce the basic techniques of Bochman’s causal calculus. Second, we will briefly contrast this rule-based logical framework with the standard structural equation modeling (SEM) approach. Finally, the main focus of the talk will be applying this calculus to formalize actual causation.
2026 Feb 26 14:00-15:30 Haoxuan Yin (尹昊萱, University of Oxford) Modal Logic as a Type System for Metaprogramming
Metaprogramming languages allow programmers to construct, manipulate and run code. In the presence of imperative features, ensuring program safety is challenging, as free variables can be passed around in references. In this talk, I present Layered Modal ML (LMML), a metaprogramming language that supports storing and running open code under a strong type safety guarantee. The type system utilises contextual modal types (Hu & Pientka, ESOP 2024) to track and reason about free variables in code explicitly. A contextual modal type \Box(\Gamma\vdash T) reads “a piece of code of type T with free variables \Gamma.” If time permits, I will also talk about using operational game semantics to model program equivalences in LMML, and how this tool can be used to verify faithfulness in metaprogramming-based program optimisations.
This is joint work with Andrzej Murawski and Luke Ong. The full version of the paper can be found at https://arxiv.org/abs/2602.03033. The talk will not suppose any prior knowledge of programming language theory, but familiarity with statically typed programming languages such as C++ or Java would be helpful.
■Past Sessions
Click HERE to check the past sessions.
Click HERE to check the past sessions from 2025.
