Artificial Intelligence (AI) and the assessment of data sensitivity in cultural organisations

该项目提出了及时的研究，以调查使用人工智能（AI）方法审查苏格兰国家图书馆各种形式的敏感性数据的有效性和适当性，重点关注遵守法律的义务（例如GDPR），但也探索更广泛的敏感信息主题。除了处理敏感信息的创新方法外，该研究还将调查更广泛的问题，例如将人工智能纳入数据处理的成本和数据伦理，并建议将人工智能用作图书馆内外敏感数据可扩展政策的组成部分。该项目旨在通过纳入人工智能来提高内存机构的工作流程生产力（例如机器学习、自然语言处理、信息检索）在数据和文献内容分析中的应用。这将涉及人工智能作为审查数据和支持敏感数据相关政策的一种手段的实验和全面探索。该项目将：1.评估和推荐可能有益的人工智能方法，不仅通过原型和/或测试人工智能性能（例如识别敏感和/或个人信息的有效性和效率），而且通过评估其解决已确定的社会问题的能力。评估社会问题，例如，与人工智能驱动的数据审查方法相关的伦理和经济风险、成本、收益和机会，作为确定或管理公众访问文档的一部分。建立超越识别敏感和/或个人信息的能力，以刺激决策和智能组织政策的制定。培养迁移学习，通过开发人工智能方法，将解决一个问题时获得的知识用于解决一个不同但相关的问题。研究保护个人权利，特别是保护弱势群体的权利，以及文件审查之间的界限，以及使用自动化设置这些界限的影响。研究问题或问题：-将人工智能支持的学习技术应用于数据和文件分析以判断其公开发布的可行性。-哪些级别需要审查（收集、对象、片段），这些如何映射到新兴政策（例如GDPR），以及需要容纳哪些类型的材料（例如图像、文本、音频、视频）？关于人工智能作为进行个人/敏感数据审查的一种手段，社会上有哪些担忧，我们如何评估他们解决这些担忧的能力？除了识别敏感数据之外，自适应的人类-AI协作对于协议开发的可行性如何？成本、限制和机会是什么？

This project proposes timely research to investigate the effectiveness and appropriateness of using artificial intelligence (AI) methods for reviewing data for diverse forms of sensitivity at the National Library of Scotland, focusing oncompliance with legal obligations (e.g. GDPR), but also exploring broader themes of sensitive information. In addition to innovative methods for handling sensitive information, the research will investigate broader concerns, such as cost and data ethics of incorporating AI in data handling, recommending guidelines for using AI as a component for scalable policies regarding sensitive data at the Library and beyond.The project aims to improve workflow productivity within memory institutions through the incorporation of AI (e.g. machine learning, natural language processing, information retrieval) in data and documentary content analysis. This will involve experimentation and a comprehensive exploration of artificial intelligence as a means of reviewing data and supporting policies regarding sensitive data.The project will:1. Evaluate and recommend AI approaches that are likely to be beneficial, not only by prototyping and/or testing AI performance (e.g. effectiveness and efficiency in identifying sensitive and/or personal information), but, by assessing their ability to address social concerns that have been identified.2. Assess social concerns, for example, ethical and economic risks, costs, benefits, and opportunities associated with AI-driven methods of data review as part of determining or managing public access to documents.3. Build capacity to go beyond identification of sensitive and/or personal information, to stimulate decision-making and intelligent organisational policy development.4. Foster transfer learning, by developing AI approaches that use knowledge gained while solving one problem to a different but related problem.5. Examine the boundaries between protection of individual rights, and particularly those of vulnerable people, and documentary censorship and the implications of using automation in setting these boundaries.Research Question or Problem:- What is the viability of applying AI enabled learning technologies, to the analysis of data and documents to adjudicate concerns about their public release.- What levels require reviewing (collection, objects, segments), how do these map to emerging policies (e.g. GDPR), and what type of material (e.g. images, text, audio, video) need accommodating?- What are the social concerns regarding AI as a means of carrying out personal/sensitive data review, and how can we evaluate their ability to address these concerns?- How feasible is adaptive human-AI collaboration for protocol development beyond identification of sensitive data? What are the costs, limitations and opportunities?