FRAIM: Framing Responsible AI Implementation and Management

FRAIM：制定负责任的人工智能实施和管理

• 批准号: AH/Z505596/1
• 负责人: Denis Newman-Griffis
• 金额: $ 31.61万
• 依托单位: University of Sheffield
• 依托单位国家: 英国
• 项目类别: Research Grant
• 财政年份: 2024
• 资助国家: 英国
• 起止时间: 2024 至 无数据
• 项目状态: 未结题
• 来源: https://gtr.ukri.org/projects?ref=AH%2FZ505596%2F1
• 关键词: FRAIM; Framing; Responsible; AI; Implementation

ContextIncreasing applications of AI technologies have necessitated rapid evolution in organisational policy and practice. However, these rapid changes have often been isolated in individual organisations and sectors, with a lack of shared cross-sectoral learning and accompanying shared values for responsible and ethical AI (RAI). Meanwhile, RAI resources have proliferated, but few effectively address the challenges in implementing shared principles, further hindering best practice development. As policymakers and industry worldwide grapple to guide, regulate, and design RAI, there is a clear need for establishing shared values and knowledge of the factors involved in implementing and managing RAI in practice.ChallengeThe Framing Responsible AI Implementation and Management (FRAIM) project will bring together cross-sector perspectives on organisational RAI policy and process to scope key stakeholders, shared values, and actionable research needs for building the evidence base on implementing and managing RAI. We have partnered with organisations representing example key areas in which AI use will significantly impact people's lives, including local policy, information access, and cultural enrichment. Our scoping work with these partners will provide a strong foundation for future development of practices and interventions to enable an ecosystem approach to RAI, and creatively and critically examine organisational implementation and management of RAI.Aims and ObjectivesThe project will focus on organisational policies and processes around using AI technologies, including pre-trained foundation models as well as context-specific machine learning, to help organise information and support decision-making. With this focus and the specific expertise of our partners, the project will address two key aims:Aim 1: Map the network of key stakeholders and values in organisational policy and process towards RAI implementation and management.Aim 2: Scope actionable needs for building a stronger evidence base around RAI implementation and management across sectors.We will achieve these aims by working with our partners to complete three objectives:Objective 1: To query and collate the values, questions, and implementation and management challenges being described in the current RAI discourse by performing a meta-analysis of RAI resources and literature.Objective 2: To identify key stakeholders, practices, and values involved in implementing and managing RAI at organisational levels by conducting exploratory interviews with staff from partner organisations.Objective 3: To scope specific plans for expanding the RAI evidence base to inform RAI implementation, management, and policy via a scoping workshop held in collaboration with project partners.Potential applications & benefitsThe scoping work in this project will establish clear directions and next steps for ecosystem-focused research and shared values to guide RAI policy and process within organisations. By drawing on multi-stakeholder perspectives and creatively engaging with the complex questions of RAI in practice, the project will benefit:Partner organisations, through shared insights from other organisations' approaches to RAI policy and process.The RAI research community, through empirically-grounded mapping of values, policies, and processes for RAI use across organisational contexts.The public, through insights into AI implementation and management and creative reflection on RAI policy and process in the world around them.

背景人工智能技术的应用不断增加，需要组织政策和实践的快速发展。然而，这些快速变化往往被孤立在各个组织和部门中，缺乏共享的跨部门学习以及负责任和道德的人工智能（RAI）的共同价值观。与此同时，RAI 资源激增，但很少能有效解决实施共享原则方面的挑战，这进一步阻碍了最佳实践的发展。随着世界各地的政策制定者和行业努力指导、监管和设计 RAI，显然需要建立共同的价值观和对实践中实施和管理 RAI 所涉及因素的了解。挑战框架负责任的人工智能实施和管理 (FRAIM) 项目将汇集有关组织 RAI 政策和流程的跨部门观点，以界定关键利益相关者、共同价值观和可行的研究需求，以建立实施和管理 RAI 的证据基础。我们与代表人工智能使用将显着影响人们生活的关键领域的组织合作，包括地方政策、信息获取和文化丰富。我们与这些合作伙伴的范围界定工作将为未来开发实践和干预措施奠定坚实的基础，以实现 RAI 的生态系统方法，并创造性地、批判性地检查 RAI 的组织实施和管理。目的和目标该项目将重点关注围绕使用人工智能技术，包括预先训练的基础模型以及特定于上下文的机器学习，来帮助组织信息和支持决策。凭借这一重点以及我们合作伙伴的具体专业知识，该项目将实现两个关键目标：目标 1：在 RAI 实施和管理的组织政策和流程中绘制关键利益相关者网络和价值观。目标 2：确定建立 RAI 实施和管理的可行需求范围围绕跨部门 RAI 实施和管理建立更强大的证据基础。我们将通过与合作伙伴合作完成三个目标来实现这些目标：目标 1：查询和整理当前 RAI 讨论中描述的价值观、问题以及实施和管理挑战通过执行对 RAI 资源和文献进行元分析。目标 2：通过对合作伙伴组织的员工进行探索性访谈，确定在组织层面实施和管理 RAI 所涉及的关键利益相关者、实践和价值观。目标 3：确定扩大 RAI 范围的具体计划RAI 证据库，通过与项目合作伙伴合作举办的范围界定研讨会，为 RAI 实施、管理和政策提供信息。潜在应用和效益 该项目中的范围界定工作将为以生态系统为中心的研究和共享价值观确定明确的方向和后续步骤指导组织内的 RAI 政策和流程。通过借鉴多方利益相关者的观点并创造性地解决 RAI 实践中的复杂问题，该项目将受益于： 合作伙伴组织，通过分享其他组织对 RAI 政策和流程方法的见解。RAI 研究社区，通过基于经验的研究绘制跨组织环境中 RAI 使用的价值观、政策和流程。公众通过深入了解 AI 实施和管理以及对其周围世界的 RAI 政策和流程进行创造性反思。