AI in Business and Society
Loading...
Paper Number
2243
Paper Type
short
Description
In the wider, public discourse on machine learning, the work of data preparation is often depicted as merely “janitorial”. Yet, emerging research shows that it involves a lot of experimentation and engaging with the domain. This is because, as data work scholars have repeatedly shown, the “embedded” nature of data requires a lot of effort, meaning making, and judgement calls to make data useful. Yet, the exact practices involved in producing training sets for machine learning are still underexplored. To address this gap, we followed five teams as they developed machine learning solutions for tackling complex, agricultural challenges. Our research finds that the development teams engage in highly creative work of using existing data to produce novel representations. In doing so, the teams engaged in three data work practices: problematizing representations, creating representational proxies, and evaluating representations and redefining a phenomenon.
Recommended Citation
Karacic, Tomislav; Günther, Wendy; Sergeeva, Anastasia; and Huysman, Marleen, "Creativity in data work: agricultural data in practice" (2021). ICIS 2021 Proceedings. 12.
https://aisel.aisnet.org/icis2021/ai_business/ai_business/12
Creativity in data work: agricultural data in practice
In the wider, public discourse on machine learning, the work of data preparation is often depicted as merely “janitorial”. Yet, emerging research shows that it involves a lot of experimentation and engaging with the domain. This is because, as data work scholars have repeatedly shown, the “embedded” nature of data requires a lot of effort, meaning making, and judgement calls to make data useful. Yet, the exact practices involved in producing training sets for machine learning are still underexplored. To address this gap, we followed five teams as they developed machine learning solutions for tackling complex, agricultural challenges. Our research finds that the development teams engage in highly creative work of using existing data to produce novel representations. In doing so, the teams engaged in three data work practices: problematizing representations, creating representational proxies, and evaluating representations and redefining a phenomenon.
When commenting on articles, please be friendly, welcoming, respectful and abide by the AIS eLibrary Discussion Thread Code of Conduct posted here.
Comments
11-AI