Pingquanqi (平权器): A Cross-Domain Sociotechnical Framework for Human-Agent Interaction Governance

Yu Wang (王宇) Independent Researcher, Shenyang, China ORCID: 0009-0002-6386-2938 [email protected] · 2026

Read Pingquanqi (平权器): A Cross-Domain Sociotechnical Framework for Human-Agent Interaction Governance by Yu Wang (王宇) Independent Researcher, Shenyang, China ORCID: 0009-0002-6386-2938 [email protected] (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2606-26573.txt.

Introduction

Yu Wang, an independent researcher in Shenyang, proposes Pingquanqi, a framework for governing how people interact with AI agents. The name means Equalizer. Wang argues that language model agents are becoming permanent infrastructure, like the electrical grid. When the chain from builders to users leaks, users pay with their lifetime, their least replaceable resource.

Pingquanqi is meant as a design standard built into agent frameworks, much as accessibility rules are built into websites. Four components are specified. A user-state model reads conversation text to level knowledge proactively. A stop-loss rule based on Bayesian reasoning caps the cost of each session. Controlled friction breaks dependency loops that feed on themselves. The stated goal is adoption as a standard, not sale as a product.

Catalog record

CREATORYu Wang (王宇) Independent Researcher, Shenyang, China ORCID: 0009-0002-6386-2938 [email protected]
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE73 KB
LICENSEhttp://creativecommons.org/licenses/by-sa/4.0/
RIGHTSCC BY SA
SOURCEarXiv

Source: https://arxiv.org/abs/2606.26573

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Pingquanqi (平权器): A Cross-Domain Sociotechnical Framework for Human-Agent Interaction Governance Yu Wang (王宇) Independent Researcher, Shenyang, China ORCID: 0009-0002-6386-2938 [email protected] Abstract Large Language Model (LLM) agents are transitioning from experimental tools to permanent infrastructure—a computational layer as enduring as the electrical grid. Like any infrastructure, they carry a cost chain: physical capital → \rightarrow enterprise investment → \rightarrow service deployment → \rightarrow user consumption → \rightarrow user lifetime. When optimized, this chain is a positive-sum loop; when unoptimized, the chain leaks, and the user’s most irreplaceable resource—lifetime—is consumed without adequate compensation. This paper proposes Pingquanqi (平权器, “Equalizer”), a cross-domain sociotechnical framework for Human-Agent Interaction Governance (HAIGF). Its technical locus lies in AI/HCI; its problem roots lie in cognitive economics; its value proposition lies in cognitive fairness. Its product form is an Agent framework-level embedded design specification—analogous to WCAG for web accessibility—whose ultimate goal is not to be purchased, but to be adopted as a standard. Pingquanqi consists of four integrated components deployable as native middleware: (1) a user-state discrimination model operating on conversation text that enables proactive knowledge leveling, (2) a Bayesian progressive stop-loss rule that caps per-session interaction cost, (3) controlled friction mechanisms that break self-reinforcing dependency loops, and (4) L steal L_{\text{steal}} , a transparency metric that renders the token-to-lifetime cost conversion visible to all parties. A fifth mechanism—reflective summarization (F5)—complements controlled friction by enabling guided cognitive recollection without disrupting cognitive quiet periods. The framework is grounded in a cross-cultural philosophical foundation: Mao’s epistemology of practice ( On Practice , 1937) provides the basis for understanding cross-session knowledge accumulation; Wang Yangming’s unity of knowledge and action (知行合一, c. 1509) illuminates the root of L steal L_{\text{steal}} —knowing without acting is incomplete knowledge, and Agent-mediated knowledge that does not transfer to user capability constitutes a form of cognitive extraction; and Hegel’s unity of theory and practice ( Philosophy of Right , 1821) demonstrates cross-traditional convergence, strengthening the argument for universal applicability. This paper argues that Pingquanqi’s primary economic beneficiary is the enterprise deploying Agent services—by reducing wasted computation, improving user satisfaction, and justifying sustained subscription revenue—with individual user benefit as the natural downstream consequence. The framework restores the positive-sum nature of the Agent infrastructure equation. Keywords: LLM agents, human-agent interaction governance, cognitive economics, token economics, cost transparency, sociotechnical framework, cross-cultural philosophy, interaction design protocol 1 Introduction 1.1 Positioning: A Cross-Domain Sociotechnical Framework Pingquanqi is not an IT product, nor is it purely an HCI contribution, nor is it an AI safety framework. It is a cross-domain sociotechnical framework —a concept with a well-established lineage ( Trist & Bamforth 1951 ) —that addresses a problem no single domain can solve alone: Table 1: Cross-domain positioning of Pingquanqi Layer Domain Pingquanqi’s Function Technical implementation AI / HCI Interaction cost measurement \rightarrow Agent framework protocol integration \rightarrow four-component engineering Problem essence Cognitive science / Behavioral economics Cognitive cost asymmetry \rightarrow token consumption quantification \rightarrow attention depletion modeling Value proposition Social philosophy / Technology ethics Cognitive fairness \rightarrow epistemological foundations \rightarrow “Completion, Not Correction” (see § 1.5) Its academic term is Human-Agent Interaction Governance Framework (HAIGF) ; its product form is an Interaction Design Protocol (IDP) embedded at the Agent framework layer; its adoption model is analogous to WCAG (Web Content Accessibility Guidelines)—a design specification whose ultimate goal is not to be purchased, but to be adopted as a standard. This positioning is deliberate. AI Fairness (with its established metrics of demographic parity and individual fairness) and Value Alignment (with its focus on RLHF and Constitutional AI) address different problem spaces. Pingquanqi addresses the incentive gradient embedded in billing models—an Agent can be safe and value-aligned while still economically incentivized to maximize user attention. This is a third pillar: economic alignment complementing safety and values. 1.2 The Agent as Infrastructure LLM Agents are not merely chatbots. They are layered systems consisting of three architectural tiers: • Model Layer. The base large language model—its architecture, training, and inference capabilities. This layer determines the quality ceiling of everything built above it. • Framework Layer. The orchestration logic that manages conversation flow, tool invocation, memory, and state. This is where interaction design decisions reside—when to respond, how to respond, whether to continue or stop. This is where Pingquanqi operates. • Infrastructure Layer. The physical compute, networking, electricity, and cooling that sustains the Agent’s operation. This layer has a non-zero, ongoing cost regardless of whether any given interaction is productive or wasteful. These three layers exist in a permanent relationship. The infrastructure layer must run continuously; the model layer must remain loaded; the framework layer must process every request. This is the permanence premise : barring catastrophic technological regression, LLM Agents are now a fixed layer of the global computational infrastructure, analogous to the electrical grid or the internet backbone. They are not regressing. 1.3 The Cost Chain Because Agent infrastructure is permanent, its costs are permanent. These costs flow through a chain of stakeholders: Physical Capital (compute, energy, cooling) \downarrow Enterprise Investment (model licensing, deployment, maintenance) \downarrow Agent Service Delivery (API or product interface) \downarrow User Payment (subscription, per-token billing, or attention-as-payment) \downarrow User Lifetime (the irreplaceable resource consumed during interaction) In an idealized configuration, this chain is a positive-sum loop : each stakeholder receives value proportional to their input, and the loop reinforces itself—satisfied users pay more, which funds better infrastructure, which delivers better service, which attracts more users. In current LLM Agent deployments, the loop is incomplete. The missing element is a mechanism that ensures the infrastructure’s consumption of user lifetime is both (a) transparent to all parties and (b) bounded by diminishing-returns logic. Without this mechanism, the infrastructure’s incentive structure—per-token billing, engagement-optimized design—drives toward maximizing token throughput regardless of whether throughput corresponds to value. 1.4 Why External Fixes Fail Attempts to correct this misalignment through external interventions—user education, regulatory mandates, timeout reminders—share a structural weakness: they fight the architecture rather than completing it. An external timeout can be ignored. A regulatory disclosure can be buried in terms of service. A user’s willpower can be eroded by design patterns optimized to erode it. The alternative, which this paper develops, is native embedding : designing cost transparency and interaction governance into the Agent framework layer itself, so that these functions are as fundamental to the Agent’s operation as memory management or tool invocation. When cost governance is native, circumventing it requires modifying the Agent’s architecture—not merely ignoring a notification. 1.5 Pingquanqi: Completion, Not Correction Pingquanqi is proposed as the missing architectural component that closes the cost chain into a positive-sum loop. It is not a post-hoc addition, a regulatory compliance feature, or a user-side tool. It is a framework-level design specification to be implemented as part of the Agent’s core orchestration logic—alongside session management, context handling, and response generation. The name 平权器 (Pingquanqi, “Equalizer”) reflects its function: it equalizes the information asymmetry between infrastructure and user by rendering costs transparent, and it equalizes the agency asymmetry by giving both parties—Agent and user—the tools to recognize when interaction has crossed the point of diminishing returns. 1.6 Contributions 1. The sociotechnical framework positioning. Pingquanqi is positioned as a cross-domain sociotechnical framework (HAIGF) with a WCAG-like adoption model, establishing that cost governance must be native rather than external

Read the complete text