Scholarlink

<p>Narratives which present smart algorithms as the major driver behind the successes of machine learning (ML) systems and ideas of automating ML development fail to acknowledge the contributions made by developers which are not directly reflected in functional code, but ground ML systems in reality. In line with ethnographic studies highlighting the importance of human collaboration and sensemaking in ML practices, we present an exercise which allows us to reflect on the consequences of reducing this dimension (Passi and Jackson in Proc ACM Hum-Comput Interact 2(CSCW):136:1–28, 2018; Neff et al. in Big Data 5(2): 85–97, 2017; Zhang et al. Proc ACM Human-Comput Interact 4(CSCW1):1–23, 2020; Muller et al. in: Proceedings of the 2019 CHI Conference on Human Factors in Computing Systems, New York, NY, USA, 2 May 2019, pp 1–5, 10.1145/3290605.3300356, 2019). In this exercise, inspired by the Surrealist artistic game, <i>cadavre exquis</i>, ML practice is re-enacted by repurposing the pipeline, a conventional ML software architecture as a purely sequential procedure, undermining collective sensemaking. We report the results of the execution of the exercise, which show that this configuration of ML practice results in a structurally coherent but semantically incoherent ML system. With this critical exercise, this paper contributes to broader scholarly discussions on the role of labor in ML and the epistemology and responsibility of data work. In addition, the exercise can be adopted to initiate discussion between critical scholars and practitioners, as well as in pedagogical contexts.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

The pipeline exquis: a critical coding exercise to re-enact ML practice

  • 期刊论文

  • 发表日期：

    2026年3月23日

查看全文
  • Juliet van Rosendaal,
  • Megan Hadasa Leal Causton

摘要

Narratives which present smart algorithms as the major driver behind the successes of machine learning (ML) systems and ideas of automating ML development fail to acknowledge the contributions made by developers which are not directly reflected in functional code, but ground ML systems in reality. In line with ethnographic studies highlighting the importance of human collaboration and sensemaking in ML practices, we present an exercise which allows us to reflect on the consequences of reducing this dimension (Passi and Jackson in Proc ACM Hum-Comput Interact 2(CSCW):136:1–28, 2018; Neff et al. in Big Data 5(2): 85–97, 2017; Zhang et al. Proc ACM Human-Comput Interact 4(CSCW1):1–23, 2020; Muller et al. in: Proceedings of the 2019 CHI Conference on Human Factors in Computing Systems, New York, NY, USA, 2 May 2019, pp 1–5, 10.1145/3290605.3300356, 2019). In this exercise, inspired by the Surrealist artistic game, cadavre exquis, ML practice is re-enacted by repurposing the pipeline, a conventional ML software architecture as a purely sequential procedure, undermining collective sensemaking. We report the results of the execution of the exercise, which show that this configuration of ML practice results in a structurally coherent but semantically incoherent ML system. With this critical exercise, this paper contributes to broader scholarly discussions on the role of labor in ML and the epistemology and responsibility of data work. In addition, the exercise can be adopted to initiate discussion between critical scholars and practitioners, as well as in pedagogical contexts.