Data-intensive curation of sustainability assessments

可持续性评估服务评估上市公司的环境、社会和治理绩效。他们的评级为金融投资者提供了重要的指导，并帮助加拿大和全球经济变得更加公正和可持续。为了产生这些数据，可持续发展分析师依赖于数据密集型自动化工作流程，这些工作流程使用人工智能（AI）技术获取和分析复杂的异构数据源。这些算法的结果需要由高技能的分析师在一种新型的数据密集型知识工作中进行解释，这种工作将人工智能与个人的专业和策展专业知识以及我们所谓的“数据密集型策展”中的判断融合在一起。由此产生的精选数据集的质量为投资相关决策提供了信息，并产生了深远的影响。这种新型的数据密集型知识工作对专家和算法之间的协作提出了挑战性的问题。分析师的专业知识和判断力与人工智能提供的分析能力一样不可替代。分析师对算法输出的信任，以及他们对其相关性的判断，取决于他们对算法输出的理解程度。由于难以追踪具体要素的来源和分类以及预测得分，分析人员很难在大海捞针中找到针，从而对具体指标进行可靠的评估。这些困难阻碍了可扩展的数据密集型管理和有效的可持续性评估。** 在这个项目中，一家全球领先的可持续发展评估公司与领先的数字策展研究人员联手解决以下问题：自动化策展机制能否帮助知识工作者更好地评估提取数据点的相关性？如何更好地支持分析师判断何时信任分类算法的结果？通过这些问题，该项目开发了数据密集型计算研究过程的概念模型，这些模型可以更好地支持下一代服务的分析和设计，从而使可持续性分析师能够进行数据密集型的策展和评估工作。

Sustainability assessment services evaluate publicly listed companies for their environmental, social and governance performance. Their ratings provide crucial guidance for financial investors and help the Canadian and global economy to become more just and sustainable. To produce them, sustainability analysts rely on data-intensive automated workflows that acquire and analyze complex, heterogeneous data sources using Artificial Intelligence (AI) techniques. The results of these algorithms need to be interpreted by highly skilled analysts in a new kind of data-intensive knowledge work that fuses AI with individuals' professional and curatorial expertise and judgment in what we call 'data-intensive curation'. The quality of the resulting curated data sets informs investment-related decision making and has far-reaching impact.**This new type of data-intensive knowledge work raises challenging questions about the collaboration between experts and algorithms. The analysts' expertise and judgment are as irreplaceable as the analytic capabilities provided by AI. The trust that analysts place in an algorithmic output, and their judgment of its relevance, rely on how well they understand where it comes from. Difficulties in tracing the origins and classifications of specific elements and predicted scores make it hard for the analysts to find the needles in the haystack that allow them to provide a robust assessment of specific indicators. These difficulties present obstacles to scalable data-intensive curation and effective sustainability assessments. **In this project, a globally leading sustainability assessment company joins forces with leading digital curation researchers to address questions such as: Can automated curation mechanisms help knowledge workers to better assess the relevance of extracted data points? How can analysts be better supported in judging when to trust the outcomes of classification algorithms? Through these questions, the project develops conceptual models of data-intensive computational research processes that better support the analysis and design of next-generation services that enable the data-intensive curation and assessment work of sustainability analysts.****