The Fourth International Workshop on Logic and Philosophy

摘要 Abstracts 

[Keynote speakers]

• Tang Liping 唐丽萍 (Sun Yat-Sen University)：
  

Title: How social connections shape linguistic politeness: combining Games, ABMs, and LLMs

Abstract: From the viewpoint of information transaction models in linguistic pragmatics, expressions of linguistic politeness (LP) impose costs upon speakers. That speakers regularly “pay” such a cost is what formal models of LP typically explain either by individual-level strategic considerations (e.g., the speaker’s aim of avoiding a face-threat to the hearer) or community-level conventional considerations (e.g., the use of LP as a relation-acknowledging device). Because these explanations are compatible, as each relates to the speaker and hearer’s social relations, we combine them into a single framework enriched by social network structures (star-shaped, ring-shaped, and fully connected networks).

We begin by constructing a basic game-theoretical model situated in social networks to capture request activities involving LP. In this model, a trade-off arises between the benefit of a successful request and the communicative cost of politeness, leading to the emergence of an optimal level of LP. To incorporate community-level effects, we define a weighted degree of LP that reflects the relative importance of strategic versus conventional considerations. Our analysis shows that the highest average optimal level of LP occurs in star-shaped networks—an outcome that intuitively aligns with power-centered, hierarchical societies.

Building on this basic game, we develop agent-based models (ABMs) and conduct simulation studies to examine the dynamics of repeated request speech acts. In scenarios with randomly occurring requests and realizations of the probabilistic responses from hearers, we observe two distinct types of communication failures, each with measurable effects on the evolution of politeness across different network structures.

Finally, moving beyond rule-based ABMs, we explore the use of LP by data-driven large language model (LLM) agents. We conduct both single-instance and repeated interaction experiments, situating LLM agents within various networks to investigate their LP behavior in request scenarios.

• Olivier Roy (Bayreuth University) :

Title: Perfect recall, knowledge memory, and action memory: individual and collective

Abstract. Perfect recall is arguably the most widely used model of individual memory in game theory and epistemic logic. This talk will begin with a historical overview to contextualise the various concepts of perfect recall that have been studied since the term was first introduced in game theory by Kuhn. Using modern tools, primarily from PDL, we will then revisit the classical decomposition of perfect recall into ‘knowledge memory’ and ‘action memory’ and provide a new characterisation of it. Next, we will transition from individual to collective memory, examining whether the components of perfect recall ‘lift’ from individual to group knowledge. In other words, we will ask whether individual perfect recall imply perfect recall defined in terms of common and distributed knowledge. Finally, we will consider what happens to these notions of collective perfect recall when group membership can be variable or uncertain. (This is joint work with Zoé Christoff (Groninen), Ondrej Majer (Prague), and Igor Sedlar (Prague).

• Liu Xiaofei 刘晓飞 (Wuhan University)：

Title: Temporal Neutrality and Future Bias

Abstract: Following Sidgwick and Rawls, many philosophers take temporal neutrality to be a basic principle of rationality. The standard conception of temporal neutrality, however, faces a well-known counterexample: future bias. Future bias is a type of bias which violates temporal neutrality yet does not appear to be irrational. Existing proposals arguing that future bias is irrational have nevertheless failed to show that the violation of temporal neutrality per se is irrational. I propose a new conception of temporal neutrality, one that explains the irrationality of future bias as a direct result of violating the principle of temporal neutrality.

• Frank Veltman (University of Amsterdam)：

Title: How `Normal’ Becomes the Norm
Abstract: At first glance, the sentence “Bears hibernate” appears purely descriptive. However, in some contexts, it can take on a normative meaning— for example, when asserted in an angry tone by a mother bear. To her cubs, it may even sound like a command to stop playing and return to the den.
    Yet, not all generic sentences lend themselves to such a normative interpretation. In my talk, I will explore which ones do, which contexts enable this shift, and how this transformation occurs. The framework of update semantics, which defines meaning not in terms of truth conditions but by its potential impact on the addressee’s intentional state, offers a powerful basis for answering these questions.
Keyword: Expectations; Defaults; Functional normativity.

[Contributed speakers] 

• Zsolt Ziegler (Budapest University of Technology and Economics)：

Title: Universalizability as Strategy: Towards a Categorical Nash Equilibrium for Social Norms

Abstract: I define Constrained Nash Equilibrium (CNE): a pre-strategic universalization test is one that screens out principles whose universal adoption would lead to a violation of viability thresholds, and on the set of admissible, a Nash equilibrium is then calculated. As usual, conditions are met, and a CNE is present.

I proved the local KKT equivalence to a prudential and price-based implementation (Self-restrained Nash, SNE) where Lagrange multipliers act as shadow costs. Thresholds (α,β) are not among the inputs of the contract but the least fixed point of a monotone operator that guarantees the best-reply stability with respect to a world model M that is common knowledge. The construction is simple and non-altruistic; it does not play a role in a one-time Prisoner’s Dilemma, however, it tightly rules the threshold settings, such as capacity, N−1 security, or carbon.

The Domain-Boundary Theorem affirms that in smooth worlds, CNE is the same as NE, whereas in threshold worlds, CNE/SNE correspond to the implementations of coherent self-restraint. Coherence comes before efficiency.

The Domain-Boundary Theorem can be explained this way. Suppose M is common knowledge, and aggregate and individual viability functions set the viability set Ω(α,β). In this context, a constitutive threshold pair is (α*,β*) which indicates that the viability set defined by it is non-empty, closed under best replies (BR-closed), and at the minimal of the partial order. The first statement is relevant to the “Smooth-World” case. If the entire strategy space S is BR-closed, then the filter is empty, and the Constrained Nash Equilibrium (CNE) is identical to the Nash Equilibrium (NE). One of the equilibrium options, KE, may be helped by its equiproportional counterfactual way if the solution is chosen amongst multiple equilibria. Besides that, under particular convexity and aggregativity assumptions, it is also capable of choosing welfare-efficient outcomes from the Nash set. The second point relates to a “Threshold-World” case. Provided that the viability set Ω(α*,β*) is non-empty and BR-closed, with α* higher than zero and β* less than one, a principle is termed an inadmissible one whose universalization causes the outcome to be out of Ω. On the resulting set of the admissible, a CNE is found. A prudential cost with player Self-restrained Nash (SNE) locally performs the output of the same KKT first-order conditions through multipliers, which serve as a cap on the implementation.

To sum it up, the forum’s survival is at the heart of the issue, so coherence really comes before efficiency. CNE is the winner in threshold worlds.

• Zhong Jiani 钟佳妮 (Trinity College Dublin)：

Title: Free Will as A Fundamentally Contested Concept

Abstract: Gregg Caruso has famously argued for a non-retributive criminal revolution. The philosophical image of free will forms the foundation of his argument, but the concept of free will is plural and contested. Here, I have two aims (1) I outline the philosophical, scientific, and folk conceptions of free will to express concern about free will sceptics who advocate a criminal justice revolution based solely on the philosophical conception. It is important to recognize that what underpins our everyday moral practices is the folk image of free will, and that the retributive punishment carried out by the state functions an outlet for the social emotions (e.g., resentment, revenge) associated with this conception. (2) Regardless of its justificatory role in retributive practices, belief in free will holds significant value in guiding human lives. I argue against the non-retributive proposal by employing a reductio argument.

Keywords: Free Will; Moral Responsibility; Criminal Justice; Retributivism; The meaning of Life

• Liu Yang 刘洋 (Unviersity of Cambridge)： 

Title: Transductive Decision Theory: Rethinking Intelligence from the Ground Up

Abstract: Classical decision theory, in its subjectivist tradition from Ramsey to de Finetti to Savage, aims to reconstruct rationality from within — grounding belief and utility in the coherent preferences of a rational agent. While the framework successfully avoids appeals to objective chance or external standards, it preserves a more elusive metaphysical commitment: the primacy of the agent. The agent is taken as an unproblematic starting point — fully individuated, preference-maximising, and epistemically closed — whose internal coherence defines the domain of rational choice.

This talk aims to encourage a conceptual reconstruction of the foundations of classical decision theory, showing that they rest not merely on structural assumptions, but on a particular image of agency. Our critique is not external or empirical, but internal and metaphysical: we examine what must be presupposed in order for the classical representation theorems to go through. In doing so, we clear the space for an alternative framework.

We propose an idea of developing a, what we call, Transductive Decision Theory (TDT) — a model of decision-making that does not presuppose the agent, but treats both agents and choices as outcomes of a process of individuation. Drawing on the philosophy of Simondon, Deleuze, and Canguilhem, we propose a transductive logic in which decisions emerge as local stabilisations within an inferential field. TDT replaces utility maximisation with the dynamics of fixpoints, and offers a new foundation for thinking about rationality, inference, and choice — one that may be better suited to emergent systems, from AI architectures to biological organisms.

• Hu Yetao 胡叶涛 (Shandong University)： 

Title: From Recombination Paradox to a Minimal Logic Core: A Paraconsistent, Non-Foundationalist Metaphysics of Modality

Abstract: This paper argues that the Recombination Paradox is an instance of Graham Priest’s Inclosure Schema. This re-diagnosis provides a powerful critical lens: conventional solutions, such as David Lewis’s ad hoc restrictions or Nolan’s appeal to proper classes, are revealed as “consistency restoration” strategies that fail to engage the paradox’s underlying structure. The paper then evaluates the dialetheic solution naturally suggested by the inclosure diagnosis. It contends that this solution is metaphysically fatal to modal realism, as by positing a possible-yet-impossible limit object, it subverts the framework’s core function of analyzing modality. The final and crucial step of the argument exposes a hidden flaw in the dialetheist’s defense: Priest’s attempt to offer a logical and metaphysical neutral account of possibility is argued to rest covertly on Logical Humean Supervenience (LHS) —a substantial, non-neutral metaphysical thesis with radical consequences of its own. Thus, the dialetheic resolution faces a profound challenge, potentially rendering it untenable for those unwilling to accept its deep metaphysical costs.

• Liu Xianrui 刘显睿 (University of Bristol)： 

Title: Logic, Thought, and Normativity: A Critique of the Constitutive-Normative Account

Abstract: In this talk I examine the contested relationship between logic and thought, with particular focus on the constitutive, normative, and constitutive-normative accounts. While the constitutive account collapses in the face of logical error and the purely normative view struggles to explain why logic holds normative authority over thought, the hybrid constitutive-normative account fails to deliver on its promise to resolve these issues, thus it remains internally inconsistent and unable to justify logic’s normative jurisdiction. Instead, I defend a version of the more traditional view: logic is constitutive of truth, while truth itself is normative for thought. Logic thus obtains normative force not directly, but via its constitutive connection to truth and the principle that thought ought to aim at truth. This layered account preserves logic’s descriptive nature while coherently grounding its normative role in human reasoning.

• Hu Yang 胡扬 (South China Normal University)：

Title: The Norm for Logical Proof: Speech Act, Natural Deduction, and the Connecting Force

Abstract: « Proof norm » indicates the compelling character of a given proof: the premises are forced to be stepwise connected and to go forward its conclusion. We address the question how such norm arises. Both Reseal (2024) and Pagin (2024) presuppose force conventionalism for answering the question, which fails to account for the natural deduction proof practice involving the inferential rules with eigenvariable conditions. I shall argue for a notion of « prover’s meaning », derived from Gricean « speaker’s meaning ». It can help us to identify what we call « conversationally illocational force » ascribed to the speech acts performed in the natural deduction proof, and it is the force commonly presumed between the prover and proofreaders. Further, such « conversationally illocational force » can further be explained as « connecting force » which gives rise to the proof norm. Last, we attempt to point out possible ways to give a formal characterization of the connecting force.

• Li Kai and Wang Yiyan 李楷/王奕岩 (Shanxi University)：

Title: Implementing Social Norms to Enhance Agency

Abstract: In this paper, we present a formal framework for norm design and evaluate proposed norms from the perspective of rule-makers. The framework assesses norms based on two criteria: implementation cost and the fairness of the resulting payoffs. Using default logic to formalize norms, along with the delta parameters proposed by Crawford and Ostrom to model payoff changes, we analyze the trade-offs between these criteria. We also examine how players’ responses to norms, as well as their expectations of others’ responses, can influence the result equilibria. Several examples illustrate how such reactions prevent a norm from achieving its intended effects.

Beyond its technical apparatus, the framework sheds light on the philosophical nature of collective agency. It shows that social groups can act as unified agents through the iterative adjustment and enforcement of norms. The formal analysis identifies the conditions under which collective intentions arise from individual compromises, and explains how normative structures both reflect and constitute shared agency. Our framework further explores the examples and consequences of norms affecting multiple strategic games, and characterizes how new norms can be introduced to correct unintended effects, if such remedification is possible.

• Yu Boning 于泊宁 (University of Maryland)：

Title: Agents’ Priors and Dynamic Networks: Two Novel Approaches to Enhancing Group Learning

Abstract: This paper investigates how two novel network arrangement strategies influence group learning within Zollman’s epistemic models. Specifically, I explore how arranging networks based on agents’ prior beliefs and implementing dynamic network structures affect a group’s accuracy and speed in reaching consensus. The explanatory framework centers on transient diversity. Results show that arrangements which initially restrict communication and then permit more information exchange significantly improve group learning effectiveness. The underlying mechanism is that such controlled communication allows agents sufficient time to independently gather evidence and develop a more cautious attitude toward belief revision, thereby reducing collective susceptibility to misleading information.

• Chen Jie 陈杰 (University of Bologna)： 

Title: Regulatory Sandboxes as Epistemic Devices: A Logical Perspective on Institutional Experimentation

Abstract: Regulatory sandboxes are increasingly used by legal institutions to manage normative and technical uncertainty in fast-evolving fields like AI and fintech. While often viewed as innovation-friendly regulatory tools, this presentation reframes them as structured epistemic devices that embody non-classical forms of reasoning under uncertainty and value conflict. Sandboxes function as meta-regulatory responses to epistemic opacity and value pluralism. They create environments where regulators, firms, and other stakeholders collaboratively explore contested concepts such as fairness, risk, and compliance. This process generates not only empirical data but also defeasible justifications for future norms, aligning with non-monotonic and argumentation-based logics. Three epistemological foundations underpin sandboxes: Experimentalism; Constructivism and Pragmatic justificationism. Formally, sandboxes can be modeled using non-monotonic logic, structured argumentation frameworks, and goal-based reasoning. These tools capture how norms and claims evolve through dynamic, multi-agent deliberation. A case study of the EU AI Act illustrates how sandboxes facilitate dialectical reasoning, where stakeholder positions are contested and refined through justification rather than rule application. This reflects a broader shift in legal governance toward defeasible, context-sensitive rationality. In conclusion, regulatory sandboxes serve not only as institutional innovations but also as real-world instantiations of logical reasoning under uncertainty, offering a fertile ground for philosophical and formal investigation.