Scholarlink

The increasing adoption of AI and autonomous systems has raised concerns about their potential negative impacts. To ensure that these systems align with human values, methodologies have been proposed to elicit SLEEC (Social, Legal, Ethical, Empathetic, and Cultural) requirements. In this paper, we present an extension of FRET, a widely used and industry-accepted tool developed by NASA, with the ability to manage SLEEC rules as formal requirements. In addition to implementing the ability to manage the lifecycle of SLEEC rules in FRET, we also provide a mechanism to generate an inference engine for SLEEC obligations and a set of run-time monitors based on Datalog. We validate our proposal on 9 use cases from the literature that include 227 SLEEC rules, demonstrating its ability to manage them. We also assess the performance and scalability of both the obligation inference engine and the monitors, demonstrating the ability of Datalog to scale to large scenarios.

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

Extending FRET with SLEEC Rules: Formalization, Obligation Inference, and Monitoring

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Mahrokh Mirani,
  • Paola Inverardi,
  • Patrizio Pelliccione,
  • Franco Raimondi,
  • Nicolas Troquard

摘要

The increasing adoption of AI and autonomous systems has raised concerns about their potential negative impacts. To ensure that these systems align with human values, methodologies have been proposed to elicit SLEEC (Social, Legal, Ethical, Empathetic, and Cultural) requirements. In this paper, we present an extension of FRET, a widely used and industry-accepted tool developed by NASA, with the ability to manage SLEEC rules as formal requirements. In addition to implementing the ability to manage the lifecycle of SLEEC rules in FRET, we also provide a mechanism to generate an inference engine for SLEEC obligations and a set of run-time monitors based on Datalog. We validate our proposal on 9 use cases from the literature that include 227 SLEEC rules, demonstrating its ability to manage them. We also assess the performance and scalability of both the obligation inference engine and the monitors, demonstrating the ability of Datalog to scale to large scenarios.