Prompted by my reading of ‘Eliciting Domain Expertise in the Absence of Formal Authority: The Case of AI Developers and Domain Experts in a Large Firm.’ AI_Developers_Domain_Experts_Formal_Authority
The early stages of designing data products, focusing especially on the role of the designers and developers of such products and their interactions with SMEs and business stakeholder technologies, host ‘upstream’ issues which may have ‘downstream’ consequences for the subsequent successful deployment and use of products. A good understanding of the reality of SME roles and the distribution of the tasks that they undertake is necessary when developers consider the viability of an idea for a data product. There are 2 aspects to be considered; firstly, that the willingness of SMEs to cooperate with developers in the short run is determined by their view as to whether the product has the potential to significantly augment the core decisions of their jobs in the long run. There is a lesser likelihood of investment in development if the focus is on decisions deemed peripheral to their everyday work.
Secondly, a robust decision analysis should be done; what is the RAPID for the decision that you are seeking to improve through the deployment of a data product? Who is really involved and accountable for success? It may not be as obvious as it first seems.
This ‘discovery’ of the decision landscape is where the role of data translator, someone who really understands the business and the potential of a data product to increase decision success adds value at the beginning of the development process. Working in collaboration with product developers, they can truly inform the likely viability of a project before it proceeds too far.
This recognition of these upstream challenges prompts a broader focus not simply on the extent to which different stakeholders might or might not adopt the right frames of reference to effectively use data products but to consider the structural (and relational) factors of authority and expertise that shape work practices surrounding the development and use of models.
