iREAL: Inclusive Requirements Elicitation for AI in Libraries to Support Respectful Management of Indigenous Knowledges

IREAL将为图书馆负责任的人工智能系统开发开发一个模式，寻求纳入土著社区的知识，特别是澳大利亚土著和托雷斯海峡岛民社区的知识。在全球范围内，图书馆拥有从“源社区”摘录的藏品；这是英国文化遗产学术中常用的一个术语，指的是土著社区。现有研究涉及与来源社区在数字化和保存土著知识方面的合作(例如，Boamah和Liew，2016年)，适应包括编目在内的专业做法(Lawther，2023年)，以及参与在国家收藏中容纳多种视角所需的“方法、伦理和实践变化”(Pringle等人，2022年)。Roke和Tillman(2022)制定了与社区接触的实用原则，以开发、描述和使用源社区收藏和知识。土著研究人员提出了“土著数据治理”(Maiam Nayri Wingara，和澳大利亚土著治理研究所2018)和以土著为中心的人工智能设计(Lewis，2020)的指导方针，但这些原则或土著社区参与在图书馆中的应用很少。因此，我们将解决在负责任的人工智能系统评估和开发方面需要明确指导的必要性，将土著权利和观点纳入作为值得信赖的人工智能的伦理基础的“尊重人类自主、防止伤害、公平和可解释性”的原则(人工智能高级别专家组，2019年)。这种指导必须具有可扩充性，因为项目资助的举措被认为不利于利益攸关方和机构之间的关系，而且是可持续的，因为处理殖民和采掘性质的收藏品的挑战影响到各种规模的机构。我们将与土著社区合作，通过与土著社区合作，扩大图书馆和其他画廊、图书馆、档案馆和博物馆(GLAM)可以利用的知识范围：将土著社区与其知识和遗产重新联系起来；促进信息管理专业人员(IMP)、研究软件工程师(RSE)和土著社区之间的关系；共同开发或批评人工智能系统，以平衡图书馆的目标与土著社区的权利和需求。我们将应用这些知识创建一个可操作的、务实的和可扩展的模型，让来源社区通过需求启发过程参与人工智能系统的评估和开发。为了实现这一目标，我们将致力于以下目标：更深入地了解土著权利和数据治理，以及它在澳大利亚和英国图书馆人工智能系统中的应用。向小鬼、RSE、土著研究人员和土著社区提供在人工智能系统中部署土著数据所面临的挑战的知识。首先基于土著和托雷斯海峡岛民社区的收藏，为图书馆人工智能系统评估或开发提出一个包容性需求的初步模型。我们将邀请来自广泛视角和国家的参与者在探索者研讨会上合作，以确定如何批评和创建负责任的人工智能系统，包括土著社区的知识和数据。鉴于这些藏书在全球范围内的广泛分布，我们的独特贡献是开发了一个需求获取过程，在考虑是否以及如何在图书馆开发或采用的系统中使用源社区数据时可以应用该过程。

iREAL will develop a model for responsible AI systems development in libraries seeking to include knowledge from Indigenous communities, specifically Aboriginal and Torres Strait Islander communities in Australia. Globally, Libraries hold collections extracted from "source communities"; a term commonly used in UK cultural heritage scholarship to refer to Indigenous communities. Existing research has addressed collaboration with source communities in digitisation and preservation of Indigenous knowledges (e.g. Boamah and Liew, 2016), adaptation of professional practices including cataloguing (Lawther, 2023), and engagement with the "methodological, ethical and practical changes" required to accommodate multiple perspectives in national collections (Pringle et al., 2022). Roke and Tillman (2022) have developed pragmatic principles for engaging with communities to develop, describe and use source community collections and knowledges.Indigenous researchers have proposed guidelines for "Indigenous data governance" (Maiam nayri Wingara, & Australian Indigenous Governance Institute 2018) and Indigenous-centred AI design (Lewis, 2020), but there has been little application of these principles, or Indigenous community engagement, in libraries. We will therefore address the need for clear guidance on responsible AI systems assessment and development that embeds Indigenous rights and perspectives into the principles of "respect for human autonomy, prevention of harm, fairness and explicability" that are the ethical basis of trustworthy AI (High Level Expert Group on Artificial Intelligence, 2019). This guidance must be scalable, as project-funded initiatives are argued to poorly serve relationships between stakeholders and institutions, and sustainable, as the challenge of dealing with collections that are colonial and extractive in nature affects institutions of all sizes. We will intervene by working with Indigenous communities to scope the knowledge by which libraries, and other Galleries, Libraries, Archives and Museums (GLAM) can: reconnect Indigenous communities with their knowledge and heritage; foster relationships between information management professionals (IMPs), research software engineers (RSEs), and Indigenous communities; and co-develop or critique AI systems to balance the objectives of libraries with the rights and needs of Indigenous communities.We will apply this knowledge to create an actionable, pragmatic and scalable model for source community engagement in the assessment and development of AI systems via the requirements elicitation process. To achieve this, we will address the following objectives:Develop a deeper understanding of Indigenous rights and data governance, and its application to AI systems in libraries in Australia and the United Kingdom.Equip IMPS, RSEs, Indigenous researchers, and Indigenous communities with knowledge of the challenges in deploying Indigenous data within AI systems.Propose a preliminary model for inclusive requirements elicitation in AI systems assessment or development for libraries, based initially upon collections from Aboriginal and Torres Strait Islander communities.We will invite participants from a broad range of perspectives and countries to collaborate in pathfinder workshops, to define how responsible AI systems might be critiqued and created including the knowledge and data of Indigenous communities. Given the widespread global distribution of these collections, our distinctive contribution is to develop a requirements elicitation process that can be applied when considering whether, and how, to use source community data in library-developed or adopted systems.