Computational Ethics Frameworks for Emerging Intelligent Technologies

Abstract

Emerging intelligent technologies -- large language models, autonomous agents, generative AI systems, biometric AI, and affective computing -- outpace the development of ethical frameworks designed to evaluate them, creating a persistent gap between technological capability and ethical governance. Traditional applied ethics frameworks derived from bioethics (beneficence, non-maleficence, autonomy, justice) provide valuable principles but insufficient operationalisation for the specific harms, power dynamics, and societal risks introduced by intelligent systems at scale. Computational ethics frameworks attempt to bridge this gap by translating ethical principles into formal structures -- mathematical representations, algorithmic procedures, and measurable criteria -- that can be applied systematically to evaluate and constrain AI system behaviour. This paper proposes the Computational Ethics Framework Taxonomy (CEFT), a systematic analysis of six computational ethics approaches -- consequentialist optimisation, deontological constraint satisfaction, virtue ethics alignment, contractualist preference aggregation, capabilities-approach evaluation, and pluralist multi-criteria frameworks -- applied to four emerging technology domains: large language models, autonomous agents, biometric AI, and affective computing. CEFT evaluates each framework across five criteria: theoretical coherence, operationalisability, conflict resolution, scalability, and cultural universality. Key results: pluralist multi-criteria frameworks achieve the highest overall applicability score (0.884); deontological constraint satisfaction best handles absolute prohibitions (privacy, consent violations); capabilities-approach evaluation is most appropriate for assessing intelligent technology impact on human flourishing. The taxonomy provides a principled framework selection guide for AI ethicists, regulators, and technology developers.