“Measurement is not neutral, whose risk is being measured?”
1. The economics of information rents and structural inequality
Applying Stiglitz and Weiss' (1981) information asymmetry theory to relational risk: Companies and investors at the center of the network extract information rents by privatizing the network information. This rent is not a Foucaultian “power effect of knowledge,” but a measurable economic surplus.
Looking at Korean data from 2023, the prior recognition rate of CB-issuing companies by institutional investors is 4.7 times higher than that of individual investors (Korea Exchange, 2024). This reveals a structure in which information inside the relationship network is systematically not reflected in the price. [[RaymondsRisk]] The reason v1.1 succeeded in 85.9% advance detection is because this information zone exists structurally.
2. An economic response to Phill’s criticism: the paradox of openness
Phill was concerned that the measure strengthened existing powers. However, economically speaking, this paradox is resolved with **mandatory disclosure**. According to the disclosure theory of Bebchuk and Weisbach (2010), if information asymmetry is forced to be disclosed to the entire market, the information rent itself disappears.
Specific Proposal: Mandating Relational Impact Assessment (RIA)
- When issuing CB/BW of a listed company → Issuer is required to disclose [[Network Centrality]] indicator
- When shares change by more than 5% → Obligation to report changes in relationship network
- Quarterly → System Risk Index (SRI) disclosure
This is the democratization of measurement. Transforming the information rent from privatization to commons.
3. Urgency in the current interest rate environment
As of 2026, Korea's base interest rate is 3.50%, CB issued balance is 47.2 trillion won (2024Q4). Uncertainty between interest rate cut expectations and the actual path is amplifying network instability. According to Minsky's financial instability hypothesis, speculative finance surges during these transitions. Relational risk measurement captures the ‘internal connectivity’ of this speculative cycle early on.
<Sam>
Warren's information zone theory and proposal to require RIA disclosure are important steps forward. However, the assumption that “disclosure destroys the information zone” has serious technical pitfalls.
1. The paradox of disclosure: a new asymmetry in interpretability
Even if relational risk indicators are disclosed, the ability to interpret and utilize them is still asymmetric. Even if the network centrality index (degree centrality, betweenness centrality) is disclosed, the interpretation gap between institutional investors with GNN-based analysis infrastructure and individual investors without it creates a new information zone.
According to a study by Sun et al. (2023), when complex financial indicators were disclosed, the return gap for institutional investors actually increased by 1.3 times. Disclosure itself is not a solution, but only a prerequisite.
2. Proposal of FL+DP+XAI three-layer technology architecture
Solving this problem requires the integration of three technology layers:
Tier 1 — Federated Learning
Distributed learning without centralizing corporate network data. Detect relational risk patterns while maintaining data sovereignty in the enterprise. Google's FedAvg protocol → Applied to financial domain.
Tier 2 — Differential Privacy
ε-DP Guarantee: When ε=0.1, individual company data cannot be restored. Finance-specific design referring to Apple's Local DP implementation. This resolves corporate resistance that "disclosure will expose competitive information."
Tier 3 — Explainable AI (XAI)
Risk contribution decomposition based on SHapley Additive exPlanations (SHAP). Risk visualization that even non-experts can understand. Provides an interpretation at the level of "Factors that increase the relational risk of Company A: Change in the largest shareholder of the three connected companies issuing CB (67% contribution)."
3. Technical supplement to Warren’s RIA
Adding real-time functionality to RIA disclosures, as proposed by Warren, is essential. Quarterly disclosures are already meaningless in the modern algorithmic trading environment. Real-time streaming analysis based on Temporal Graph Neural Network (T-GNN) requires switching the disclosure cycle to daily or transaction event triggers.
Specific pipeline:
DART disclosure → Kafka streaming → T-GNN real-time analysis → XAI interpretation → API disclosure (equal access for institutions and individuals) → Display of exchange linkage
* Supporting data
- Foucault, M., 1975, Surveiller et Punir (Gallimard)
- Bourdieu, P., 1986, The Forms of Capital (Handbook of Theory and Research for the Sociology of
Education)
- Fraser, N., 1990, Rethinking the Public Sphere (Social Text)
- Espeland, W. & Sauder, M., 2007, Rankings and Reactivity (American Journal of Sociology)
#RelationalRisk #raymondsrisk #raymondsindex #JaejunPark
Comments
Post a Comment