AI Microservices for Sustainable Innovation in Education

Automation in education is increasingly applied to processes requiring pedagogical judgement. We argue that such automation should be separate from general platforms used to administrate learning and teaching activities. We propose an architecture for the wider ecosystem based on platforms calling external microservices to provide pedagogical judgements, and that this approach will enable innovation and foster ethical approaches. We provide a theoretical framework for microservices based on a literature review of software architecture principles and a characteristion of the education sector. We then present results of practical deployment of an exemplar system for providing automated formative feedback. The system delegates all pedagogical judgement to external microservices. The system was deployed to thousands of active student users, covering 13,000 study tasks and 1.5 million calls to external services. We present case studies of the general deployment, and two specific cases of automated feedback on university-level mathematical proofs, and school level essays in English language. The case studies on deployments provide large scale evidence for the feasibility of microservices to provide pedagogical feedback. The evidence presented also supports the theoretical framework that microservices for pedagogical judgment promote innovation from a wide variety of experts in niche areas, emancipate the services to use technology relevant to the specific problem, prevent platform lock-in leading to pedagogy lock-in, and foster transparency, equity of access, and higher quality services. We discuss the future prospects of a microservice architecture for the education ecosystem, including the potential risks of such an approach and future research needs. Our main conclusion is that we present a framework for a new approach to automating pedagogical judgments, and novel, large scale, practical evidence for its feasibility.