H.Wanhong

Open Question

1. Can the attribute family ${X_i}$ be endogenised — derived from the
   system’s own dynamics rather than imposed externally?

2. If social systems are non-ergodic, what replaces the equilibrium as the
   object of normative analysis? Is justice a topological invariant of the
   trajectory rather than a property of any state?

3. Is the belief state $b(s)$ in a POMDP framework the correct formalisation
   of the epistemic suspension characteristic of boundary states — social facts
   that are simultaneously valid and invalid?

4. Can generative justice be formalised as a functional $J: \Omega^\mathbb{T}
   \to \mathbb{R}$ over the space of trajectories, such that some paths are more
   just than others independent of their endpoints?

5. What is the universal grammar underlying the semiotic exchange across
   incommensurable symbol systems — in forests, in markets, in intimate
   relationships?

Miranda Fricker formally introduced the concept of “epistemic injustice” in her 2007 monograph of the same name, identifying a distinctive form of wrongful treatment: the denigration or deprivation of another person’s capacity and standing as a knower. Unlike traditional ethical frameworks that focus on harm to agents as actors, epistemic injustice attends to the systematic exclusion individuals face in the production, transmission, and interpretation of knowledge. This article offers a structured overview of the concept’s characteristics, taxonomy, harm structure, and paths to redress, drawing on the primary literature and related scholarship.

I. Two Basic Forms of Epistemic Injustice

1. Testimonial Injustice

Testimonial injustice occurs when a hearer assigns a speaker’s testimony less credibility than it deserves, with the deficit caused by identity prejudice operating in the hearer. Such prejudice is typically not deliberate; it is rooted in historically sedimented structures of social power and infiltrates epistemic assessment through the mediation of stereotypes related to race, gender, class, and other identity markers.

From the standpoint of epistemic norms, testimonial injustice constitutes a violation of the evidentialist duty: if the evidence supports the probability of a proposition P being true above a reasonable threshold, a rational agent ought to believe P. When prejudice intervenes in this process, epistemic judgment is calibrated not to evidence but to social signifiers, and the epistemic norm is thereby undermined.

At the level of consequences, testimonial injustice not only directly diminishes the speaker’s epistemic agency but also disrupts the legitimate circulation of knowledge and truth in society, producing a structural rupture in the chain of epistemic transmission.

Background

Argumentative writing is a core academic and civic competency, requiring learners to formulate claims, support them with evidence, and articulate coherent reasoning. Automated Essay Scoring (AES) has long been proposed as a scalable alternative to manual scoring, and the emergence of LLMs has dramatically accelerated this field. Early AES systems focused on general writing quality rather than argumentative reasoning, and prior reviews have not addressed LLM-based scoring of argumentative essays specifically. The field has grown rapidly but without consolidated methodological, psychometric, or ethical foundations.

Motivation

Despite rapid growth, LLM-based AAES remains conceptually unsettled. Scoring rubrics in existing datasets disproportionately emphasize rhetorical and linguistic fluency while neglecting deeper argumentative constructs such as logical cogency, evidential sufficiency, and dialectical engagement. Critical concerns persist around reliability, construct validity, fairness, and responsible deployment — particularly given the high-stakes educational contexts in which such systems may be used. No prior systematic or scoping review had addressed these gaps comprehensively.

Research Questions

• RQ1: How are LLMs currently employed for automated scoring of argumentative essays and feedback provision in educational settings — what techniques, datasets, and evaluation methodologies are used, and what methodological gaps remain?
• RQ2: To what extent do LLM-based AAES approaches align with human judgment in terms of psychometric validity and the FATEN principles (Fairness, Augmentation, Transparency, bEneficence, Non-maleficence) for responsible educational assessment?

Wanhong HUANG
April 2026

Abstract

Policy reports occupy a distinctive epistemic position between empirical research and administrative action. Unlike academic papers, which culminate in findings, policy reports culminate in recommendations — they are instruments of directed change. Yet the field lacks a consolidated methodological grammar for how such reports should be structured, what types of evidence they should marshal, and how their evaluative logic should be justified. This essay proposes a two-type taxonomy of policy research orientations, articulates the internal architecture of a rigorous policy report, and discusses the epistemological standards that lend such documents their persuasive force. A curated reference list of authoritative document repositories is provided as an appendix.

1. Introduction

The genre of the policy report is, structurally speaking, an argument. Its reader — a minister, a programme director, an institutional board — arrives with a practical question: what should we do? Answering this question well requires the author to establish not only what is true about the present state of affairs, but also what is normatively at stake and what interventions are operationally feasible. This tripartite burden — descriptive, normative, and operational — distinguishes the policy report from the academic article, the audit report, and the journalistic investigation, each of which bears only a partial version of it.

The core narrative logic of a well-constructed policy report can be stated simply: why this mattered, what we did, what we found, what you should do. Every structural component of the document should serve one of these four moments. Components that serve none of them are not merely superfluous; they actively dilute the report’s persuasive force.

This essay proceeds as follows. Section 2 introduces a taxonomy of policy research orientations. Section 3 proposes a canonical component architecture for the policy report. Section 4 addresses the epistemological basis of persuasive state inference. Section 5 discusses evaluation standards and quality assurance. Section 6 concludes with observations on the relationship between structure and institutional legitimacy.

PDF Source (PDF File Link)

中文

各位老师、各位同事，大家周一早上好！

今年，樱花开放的季节又一次到来了。
樱花真是十分美丽。

樱花之美，不仅体现在它的色彩与姿态之中，也蕴藏于其深邃的“结构”之中。

枝条如何分岔，花朵如何簇集，而当微风吹起之时，原本维持着的花之集合又会被动态地解体，并在空间之中重新组织为新的相态。
这些现象都仿佛在提醒我们：在不断变化的表象之下，某些本质性的结构关系仍然得以保持，而在临界点上，又可能发生剧烈的拓扑相变。

在连续变形之中依然保持不变的性质，以及在越过某个阈值瞬间所产生的结构性飞跃，
正是拓扑学与动力学相交之处最核心的研究主题。

若将这一视角进一步推进，EEG 或许也可以被理解为一种拓扑—动力学的演化过程。
在复杂的脑动态之中，那些在平常状态下稳健保持的拓扑不变量，会如何在癫痫发作之前崩解，或者转移为新的结构。
若能够捕捉这样的临界征兆，便有可能为癫痫预测提供一种极具潜力的描述框架。

因此，我在下方列出了一些关于利用拓扑方法研究 EEG 分析与动力学的论文 ([1-5])，并附上相关资料。
若能对大家稍有帮助，我将十分欣慰。

值此樱花盛开的时节，也祝愿各位都能好好享受这美好的春日时光。

黄万鸿
2026年3月30日

参考文献

[1] Chrétien, S., Gao, B., Thebault-Guiochon, A., & Vaucher, R. Time topological analysis of EEG using signature theory.
[2] Gaurav, K., Sharma, N., Landge, J., & Bollu, T. K. R. Mapping EEG Sensor Networks: Persistent Topology-Driven Learning for Affective States Recognition. IEEE.
[3] Ling, C. Y.-F., Phang, P., & Liew, S.-H. Topological data analysis in EEG signal processing: a review.
[4] Döner, B., Ingolfsson, T. M., Benini, L., & Li, Y. LUNA: Efficient and Topology-Agnostic Foundation Model for EEG Signal Analysis.
[5] Xi, X., Fan, Z., Wang, T., Li, L., & Yang, J. (2025). Topology analysis of EEG-based functional brain network after stroke. Neurocomputing, 637(C).

English

Good Monday morning, everyone!

Once again, the season of cherry blossoms has come around.
The cherry blossoms were beautiful.

The beauty of cherry blossoms lies not only in their color and appearance, but also in their profound underlying structure.

The branching of the limbs, the clustering of the blossoms, and then, when the wind begins to blow, the once-stable assemblage of flowers is dynamically dismantled and reconfigured into a new phase in space.
All of these phenomena suggest that beneath ever-changing appearances, certain essential structural relations are preserved, while at critical points, dramatic topological transitions may occur.

Properties that remain invariant through continuous deformation, and structural leaps that emerge the moment a threshold is crossed:
this is precisely the core of inquiry where topology and dynamics intersect.

From this perspective, EEG as well may be understood as a process of topological-dynamical transformation.
Within the complex dynamics of the brain, one may ask how the topological invariants that are ordinarily maintained with robustness begin to collapse, or transition into new structures, immediately before an epileptic seizure.
Capturing such critical signs could provide a highly promising descriptive framework for seizure prediction.

Accordingly, I have listed below several papers ([1-5]) concerning EEG analysis and dynamics using topological methods, and have attached the related materials.
I would be very happy if they prove even slightly helpful.

In this season when the cherry blossoms are in full bloom, I also wish all of you a wonderful spring.

Wanhong Huang
March 30, 2026

References

[1] Chrétien, S., Gao, B., Thebault-Guiochon, A., & Vaucher, R. Time topological analysis of EEG using signature theory.
[2] Gaurav, K., Sharma, N., Landge, J., & Bollu, T. K. R. Mapping EEG Sensor Networks: Persistent Topology-Driven Learning for Affective States Recognition. IEEE.
[3] Ling, C. Y.-F., Phang, P., & Liew, S.-H. Topological data analysis in EEG signal processing: a review.
[4] Döner, B., Ingolfsson, T. M., Benini, L., & Li, Y. LUNA: Efficient and Topology-Agnostic Foundation Model for EEG Signal Analysis.
[5] Xi, X., Fan, Z., Wang, T., Li, L., & Yang, J. (2025). Topology analysis of EEG-based functional brain network after stroke. Neurocomputing, 637(C).