Responsible AI for Long-term Trustworthy Autonomous Systems (RAILS): Integrating Responsible AI and Socio-legal Governance

用于长期可信自治系统（RAILS）的负责任的人工智能：将负责任的人工智能与社会法律治理相结合

• 批准号: EP/W011344/1
• 负责人: Lars Kunze
• 金额: $ 90.48万
• 依托单位: University of Oxford
• 依托单位国家: 英国
• 项目类别: Research Grant
• 财政年份: 2022
• 资助国家: 英国
• 起止时间: 2022 至 无数据
• 项目状态: 未结题
• 来源: https://gtr.ukri.org/projects?ref=EP%2FW011344%2F1
• 关键词: Responsible; AI; Long; term; Trustworthy

Society is seeing enormous growth in the development and implementation of autonomous systems, which can offer significant benefits to citizens, communities, and businesses. The potential for improvements in societal wellbeing is substantial. However, this positive potential is balanced by a similar potential for societal harm through contingent effects such as the environmental footprint of autonomous systems, systemic disadvantage for some socio-economic groups, and entrenchment of digital divides. The rollout of autonomous systems must therefore be addressed with responsibilities to society in mind. This must include engaging in dialogue with society and with those affected, trying to anticipate challenges before they occur, and responding to them. One such anticipated challenge is the effect of change on autonomous systems. Autonomous systems are not designed to be deployed in conditions of perfect stasis, as they are unlikely to encounter such conditions in real-world environments. They are frequently designed for changing environments, like public roads, and may also be designed to change themselves over time, for instance by means of learning capabilities. Not only that, but these changes in deployed systems and in their operating conditions are also likely to take place against a shifting contextual background of societal alteration (e.g. other technologies, 'black swan' events, or simply the day-to-day operation of communities). The effects of such change, on the systems themselves, on the environments within which they are operating, and on the humans with which they engage, must be considered as part of a responsible innovation approach. The RAILS project brings together a team from UCL and the Universities of York, Leeds and Oxford, from multiple disciplines, with the aim of engaging with the challenges associated with the long-term operation of autonomous systems and the effects of change on these systems. In particular, we will explore how the notion of responsibility is affected by (i) open-ended dynamic environments - situations that change over time, and(ii) lifelong-learning systems - i.e. systems that are designed to adapt themselves to their circumstances and 'learn' over time. The RAILS project will focus on such independent long-term autonomous systems in different applications. These will include (i) autonomous vehicles and (ii) autonomous robot systems such as unmanned aerial vehicles (drones). RAILS will look at social and legal contexts, as well as technical requirements, in order to assess whether and how these systems can be designed, developed, and operated in a way that they are responsible, accountable, and trustworthy. The overall aim of the RAILS project is to bring together responsible development principles with governance mechanisms and technical understanding to create new understandings of how autonomous systems can adapt to change, how they can be deployed in a responsible and trustworthy way, and how such deployment can be framed by governance to ensure accountability and flexibility.

社会正在看到自治系统的开发和实施的巨大增长，这可以为公民，社区和企业带来重大利益。改善社会福祉的潜力巨大。然而，这种积极的潜力被类似的潜在社会危害所抵消，这些潜在社会危害是通过诸如自治系统的环境足迹、某些社会经济群体的系统性劣势以及数字鸿沟的加深等偶然影响造成的。因此，在推出自主系统时必须考虑到对社会的责任。这必须包括与社会和受影响者进行对话，努力在挑战发生之前预测挑战，并应对挑战。其中一个预期的挑战是变化对自治系统的影响。自主系统的设计并不是为了在完美停滞的条件下部署，因为它们不太可能在现实世界的环境中遇到这样的条件。它们通常是为不断变化的环境而设计的，如公共道路，也可能是为了随着时间的推移而改变自己，例如通过学习能力。不仅如此，部署的系统及其运行条件的这些变化也可能发生在社会变革的不断变化的背景下（例如，其他技术，“黑天鹅”事件，或者只是社区的日常运作）。这种变化对系统本身、系统运行的环境以及系统所涉及的人的影响，必须作为负责任的创新方法的一部分加以考虑。RAILS项目汇集了来自伦敦大学学院、约克大学、利兹大学和牛津大学的多个学科的团队，旨在应对与自主系统长期运行相关的挑战以及变化对这些系统的影响。特别是，我们将探讨责任的概念是如何受到（i）开放式动态环境-随着时间的推移而变化的情况，以及（ii）终身学习系统-即旨在适应其环境并随着时间的推移而“学习”的系统。RAILS项目将专注于不同应用中的独立长期自主系统。这些将包括（i）自动驾驶车辆和（ii）自动机器人系统，如无人驾驶飞行器（无人机）。RAILS将研究社会和法律的背景以及技术要求，以评估这些系统是否以及如何以负责任、负责任和值得信赖的方式设计、开发和运行。RAILS项目的总体目标是将负责任的开发原则与治理机制和技术理解结合在一起，以创建对自治系统如何适应变化的新理解，如何以负责任和值得信赖的方式部署它们，以及如何通过治理来构建这种部署以确保问责制和灵活性。

项目成果

Simulation-Based Counterfactual Causal Discovery on Real World Driver Behaviour

基于模拟的现实世界驾驶员行为的反事实因果发现

• DOI: 10.1109/iv55152.2023.10186705
• 发表时间: 2023
• 作者: Howard R

Evaluating Temporal Observation-Based Causal Discovery Techniques Applied to Road Driver Behaviour

• DOI: 10.48550/arxiv.2302.00064
• 发表时间: 2023-01
• 期刊: ArXiv
• 作者: Rhys Howard;L. Kunze

Explainable Action Prediction through Self-Supervision on Scene Graphs

• DOI: 10.1109/icra48891.2023.10161132
• 发表时间: 2023-02
• 期刊: 2023 IEEE International Conference on Robotics and Automation (ICRA)
• 作者: Pawit Kochakarn;D. Martini;Daniel Omeiza;L. Kunze

CC-SGG: Corner Case Scenario Generation using Learned Scene Graphs

• DOI: 10.48550/arxiv.2309.09844
• 发表时间: 2023-09
• 期刊: ArXiv
• 作者: George Drayson;Efimia Panagiotaki;Daniel Omeiza;Lars Kunze

The SAGE Handbook of Digital Society

数字社会 SAGE 手册

• DOI: 10.4135/9781529783193
• 发表时间: 2023
• 作者: Housley W