摘要
In this paper, we explore the questions "How can we make the impact of the data scientists' subjectivity on the design of machine learning (ML) models more transparent?". We also argue that it is insufficient to only focus on the technology of ML. We should also design the broader socio-technical system within which the ML model is to be deployed. We draw on philosophers and systems thinkers to offer two additional steps to current ML design process models. These new phases focus on a) understanding the subjectivity of the data scientists involved in ML design and b) holistically examining the wider social context within which the ML model is to be implemented. We provide insight into how to implement the additional steps in the ML design process. The proposed additions help address issues such as the lack of explainability/transparency in ML models, bias in ML, and difficulty in managing human-ML hybrids.
学科与关键词
- Toward Sociotechnical AI: Mapping Vulnerabilities for Machine Learning in ContextDobbe, Roel*; Wolters, AnoukMINDS AND MACHINES, 2024, 34(2): 12.被引 1
- WHEN THE MACHINE MEETS THE EXPERT: AN ETHNOGRAPHY OF DEVELOPING AI FOR HIRINGvan den Broek, Elmira*; Sergeeva, Anastasia; Huysman, MarleenMIS QUARTERLY, 2021, 45(3): 1557-1580.被引 124
- Towards Sustainability of AI - Identifying Design Patterns for Sustainable Machine Learning DevelopmentLeuthe, Daniel*; Meyer-Hollatz, Tim; Plank, Tobias; Senkmueller, AnjaINFORMATION SYSTEMS FRONTIERS, 2024.被引 1
