Machine learning (ML) offers widely-recognized, but complex, opportunities for both public and private sector organizations to generate value from data. A key requirement is that organizations must find ways to develop new knowledge by merging crucial ‘domain knowledge’ of experts in relevant fields with ‘machine knowledge’, i.e., data that can be used to inform predictive models. In this paper, we argue that understanding the process of generating such knowledge is essential to strategically develop ML. In efforts to contribute to such understanding, we examine the generation of new knowledge from domain knowledge through ML via an exploratory study of two cases in the Swedish public sector. The findings reveal the roles of three mechanisms – dubbed consolidation, algorithmic mediation, and naturalization – in tying domain knowledge to machine knowledge. The study contributes a theory of knowledge production related to organizational use of ML, with important implications for its strategic governance, particularly in the public sector.
Highlights
To leverage the benefits of machine learning (ML), organizations need to effectively generate relevant knowledge from their data.
We present an exploratory case study involving two ML initiatives in the Swedish public sector, examining how new knowledge is generated from domain knowledge through ML.
We introduce a theoretical model that elucidates the connection between domain knowledge and machine knowledge during the development of ML systems.
We indicate important strategic implications for governing ML in public sector organizations.
תגובות