Scholarlink

<p>Artificial intelligence systems increasingly function as structural interventions within ecological, social, and cognitive systems, producing effects that propagate across temporal and spatial scales. Existing ethical frameworks—ranging from utilitarian optimisation to deontological constraint systems—lack a unifying criterion capable of evaluating these system-level, long-horizon impacts. As a result, contemporary AI governance remains vulnerable to abstract proxy optimisation failures, underdetermined trade-offs, and the absence of principled intervention thresholds. This article introduces the Qualitative Model of Ethics (QME), a naturalistic and teleological framework grounded in a single evaluative scalar: the generative capacity of the Whole Living System (WLS). Generative capacity is defined as the ability of a system to produce, sustain, and diversify measurable structure across time. Ethical valence is identified with directional changes in this capacity, formalised as <i>GC</i><sub><i>WLS</i></sub>[<i>t</i><sub><i>0</i></sub>, <i>t</i><sub><i>H</i></sub>]. Rather than attempting to compute total potential future content, the model evaluates the conditions under which future generativity remains accessible, including ecological stability, population viability, and system optionality. The Qualitative Ethics Stack (QES) operationalises this criterion through a hierarchical indicator structure and explicit inhibition thresholds, enabling evaluation under conditions of partial observability and uncertainty. The framework is demonstrated across heterogeneous cases, including solar radiation management, engagement-optimising recommender systems, and the collapse of the Aral Sea ecosystem. QME departs from ontocentric and universalist approaches such as Floridi’s information ethics by adopting a conditionally valenced and teleological structure grounded in measurable system dynamics. It defines harm not as entropy in a metaphorical sense, but as reduction in system-level generative capacity, including cases of pathological production where local optimisation degrades global viability. The model is explicitly falsifiable and designed for interdisciplinary application. By grounding ethical evaluation in the conditions that enable the continued generation of measurable structure, QME provides a unified and operational framework for AI governance at scale.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

A qualitative model of ethics: a conceptual overview and architectural implications for AI governance 2026

  • 期刊论文

  • 发表日期：

    2026年7月8日

查看全文
  • Joel Praharenka

摘要

Artificial intelligence systems increasingly function as structural interventions within ecological, social, and cognitive systems, producing effects that propagate across temporal and spatial scales. Existing ethical frameworks—ranging from utilitarian optimisation to deontological constraint systems—lack a unifying criterion capable of evaluating these system-level, long-horizon impacts. As a result, contemporary AI governance remains vulnerable to abstract proxy optimisation failures, underdetermined trade-offs, and the absence of principled intervention thresholds. This article introduces the Qualitative Model of Ethics (QME), a naturalistic and teleological framework grounded in a single evaluative scalar: the generative capacity of the Whole Living System (WLS). Generative capacity is defined as the ability of a system to produce, sustain, and diversify measurable structure across time. Ethical valence is identified with directional changes in this capacity, formalised as GCWLS[t0, tH]. Rather than attempting to compute total potential future content, the model evaluates the conditions under which future generativity remains accessible, including ecological stability, population viability, and system optionality. The Qualitative Ethics Stack (QES) operationalises this criterion through a hierarchical indicator structure and explicit inhibition thresholds, enabling evaluation under conditions of partial observability and uncertainty. The framework is demonstrated across heterogeneous cases, including solar radiation management, engagement-optimising recommender systems, and the collapse of the Aral Sea ecosystem. QME departs from ontocentric and universalist approaches such as Floridi’s information ethics by adopting a conditionally valenced and teleological structure grounded in measurable system dynamics. It defines harm not as entropy in a metaphorical sense, but as reduction in system-level generative capacity, including cases of pathological production where local optimisation degrades global viability. The model is explicitly falsifiable and designed for interdisciplinary application. By grounding ethical evaluation in the conditions that enable the continued generation of measurable structure, QME provides a unified and operational framework for AI governance at scale.