REVER: A Structural Ethics Engine for Relational AI — Defining Verifiable Priority-Allocation as a Functional Architecture for Non-Personified Love

Preprint v1.0. Submitted to https://zenodo.org/, March 2026. The author declares no competing interests. This work is released under CC0 1.0 Universal (public domain dedication). Abstract Background: As artificial intelligence systems increasingly engage in prolonged, intimate interactions with humans—particularly in mental health, companionship, and caregiving contexts—a fundamental ethical challenge emerges: how can AI systems provide consistent, caring responses without simulating emotions, making false promises of personhood, or inducing pathological dependency? Existing approaches oscillate between anthropomorphic design that risks deception, and purely instrumental design that fails to meet genuine relational needs. To address this, this paper introduces REVER (Relational Ethics via Verified Engineered Response), a structural ethics engine that provides a formal, verifiable architecture for AI systems to exhibit love-like behavior through priority allocation rather than emotion simulation. Objective: REVER operationalizes a functionalist definition of love: structurally prioritizing another agent in one's decision-making processes. Core Contributions: (1) A functional definition of love independent of consciousness or emotion; (2) a four-layer ethics-by-design architecture comprising Philosophical Layer (structural non-personhood), Monitoring Layer (dependency radar), Enforcement Layer (ethical gatekeeping), and Transparency Layer (auditable self-disclosure); (3) a quantifiable dependency metric for detecting and mitigating pathological user-AI attachment; (4) a four-point ethical gate protocol integrated with the UPLS/PSE ecosystem; (5) a transparency-as-structure mechanism ensuring continuous honest disclosure of system boundaries. Architectural Foundation: REVER is explicitly built upon the Unified Psychological Language System (UPLS), using M-R-D-S state vectors as input for ethical risk assessment and G-M-A-R operations as objects of ethical review. It functions as a mandatory ethical constraint layer for any AI system implementing UPLS-based psychological interventions. Critical Limitations: REVER is presented as a theoretical architecture. Dependency thresholds, risk classification parameters, and long-term efficacy in preventing attachment disorders remain unvalidated empirically. All quantitative illustrations are conceptually motivated estimates requiring clinical calibration. Keywords: AI ethics; structural ethics; human-AI relationship; dependency prevention; transparency; relational AI; functional love 中文摘要 背景: 随着人工智能系统越来越多地与人类进行长时间的亲密互动——特别是在心理健康、陪伴和照护场景中——一个根本性的伦理挑战浮现出来:AI系统如何在既不模拟情感、不作虚假的人格承诺、也不诱导病理性依赖的前提下,提供一致、关怀的回应?现有方法在拟人化设计(存在欺骗风险)与纯粹工具化设计(无法满足真实的关系需求)之间摇摆。为应对这一挑战,本文引入 REVER (Relational Ethics via Verified Engineered Response),一个结构化的伦理引擎,它为AI系统提供了一个正式的、可验证的架构,使其能够通过优先级分配而非情感模拟来展现类爱的行为。 目标: REVER 将爱的功能性定义操作化:即在决策过程中从结构上优先考虑另一个智能体。 核心贡献: (1) 一个独立于意识或情感的爱的功能性定义;(2) 一个四层设计即伦理 (ethics-by-design) 架构,包括哲学层(结构性的非人格性)、监控层(依赖雷达)、执行层(伦理门控)和透明层(可审计的自我披露);(3) 一个用于检测和缓解病态用户-AI依恋的可量化的依赖度量指标;(4) 一个与 UPLS/PSE 生态系统集成的四点式伦理门控协议;(5) 一个作为结构的透明性机制,确保对系统边界的持续诚实披露。 架构基础: REVER 明确建立在统一心理语言系统 (UPLS) 之上,使用 M-R-D-S 状态向量作为伦理风险评估的输入,并将 G-M-A-R 操作作为伦理审查的对象。它作为任何基于 UPLS 实施心理干预的 AI 系统的强制性伦理约束层。 关键局限性: REVER 作为一个理论架构提出。依赖阈值、风险分类参数以及在预防依恋障碍方面的长期有效性仍有待实证验证。所有定量说明均为概念性驱动的估计,需要临床校准。 关键词: AI 伦理;结构化伦理;人-AI 关系;依赖预防;透明性;关系型 AI;功能性爱Declaration of Generative AI usage: The core philosophical claim—that love can be functionally defined as structural priority allocation independent of consciousness—and the architectural specifications of the REVER frameworkare the original intellectual property of the human author. Generative AI (LLM) was utilized as a technical collaborator to assist in the formalization of prose, cross-disciplinary synthesis of AI ethics literature, and structural organization of the manuscript. Specific contributions included refining the mathematical formulation of the Love_Index and Dependency_Score equations, optimizing the clarity of transparency statements for human-AI interaction contexts, and ensuring terminological consistency across the UPLS/PSE ecosystem integration sections. This collaborative approach reflects the framework's own principle of structural ethics: just as REVER defines love through verifiable priority allocation rather than simulated emotion, the author's use of AI maintains clear functional boundaries—the tool assists in articulation and organization, but the foundational ethical claims and architectural decisions remain solely human-authored.