Discourses on AI in electricity systems: technology assessment narratives in academic and policy documents

This paper focuses on technology assessment narratives (TANs) of academic and policy actors regarding artificial intelligence (AI) in the Dutch electricity system. TANs discuss expectations concerning the use and sociotechnical impact of a technology. We map and compare the TANs apparent within European Union and Dutch policy documents (the policy domain) and technology-focused academic articles (the technology domain) between 2000 and 2020 regarding the implementation of AI in the electricity sector. For this, we develop a method that allows us to compare the core elements of TANs across domains using a framework of six categories of analysis: (1) technological capabilities of AI, (2) level of sociotechnical disruption, (3) area of application of AI (sub-system), (4) innovation phase, (5) policy area, and (6) policymaking phase. We find that actors in the policy domain focus mainly on AI with autonomous decision-making capabilities, and AI implementations which would have a disruptive sociotechnical impact, whereas actors in academia also consider AI without decision-making capabilities and implementations with an incremental sociotechnical impact. The policy domain therefore takes a generic approach when it comes to the governance of AI, which may fall short when determining policy for specific AI applications.