The social life of models: Studying the makings of digital models and questions of code-ability

What goes into creating a digital model? The selection and quality of training data is generally emphasized, and anthropologists tend to broaden the analysis by considering relational, political, economic, and value-related factors that contribute to shaping the conception of a digital model (Seaver 2017; Suchman 2023). The imaginaries and envisioning of the problems to be solved, along with the human expertise to be assisted or replaced also warrant inquiry. But certain tacit professional skills pertaining to carrying out the tasks are often overlooked. This may stem from a perception that these skills are irrelevant or simply that they cannot be translated into code - they are not ""code-able"" (Amoore & Piotukh 2015). This shapes not only the technologies but transforms how societal problems are reconfigured as machine learning problems (Amoore 2023).
In this presentation, I introduce a newly launched project involving engineers, shipowners, and suppliers collaborating to improve energy efficiency and reducing carbon emissions in shipping through the development of mathematical models and digital twins. Conceiving of the ""user"" as everyone involved and ""the human factor"" as everything contributing to the process, I present two central questions that I’m currently grappling with: For one, which forms of expertise and often tacit sensory skills fall out of the mathematical models, either because they are not acknowledged or because they are not seen as within the realm of code-ability? And secondly, how can we study the makings of a digital model ethnographically?