It has been recognized by many leaders in the open education community that the rise of adaptive, personalized and otherwise data-driven approaches to leveraging digital learning content have profound implications for OER. If open educational resources continue to be focused on the development of relatively static, textbook-like materials that are unable to engage with data-driven feedback loops, then the materials developed by closed approaches will rapidly outpace OER with regard to effectiveness and impact. This state of affairs will likely result in OER being relegated to second-class status, used by disadvantaged learners who cannot afford high-tech, data-driven courseware.

Ensuring that OER can continue to support effective, data-driven learning experiences will require progress on a number of fronts with regard to the open infrastructure – open materials that can generate actionable data in relation to learner achievement and knowledge state; open systems to serve these materials and capture necessary information about learner interactions; open data services, equipped to consume data streams from interactions with open courseware and return processed information regarding knowledge state, achievement, progression and engagement, among other components. While some significant elements of this ecosystem are missing, the open community continues to make progress in building out components of open analytics systems. Unfortunately, too often the 'open' of these systems is synonymous with open source software - open code that can consume and analyze data feeds - without exploring the implications of the underlying models and algorithms that process the data and return interpretable information about the learners, their outcome achievements, risks, progress and attainment. While the openness of the supporting systems and course materials is important, over the longer term ensuring that the underlying analytic methods, models and algorithms are open is essential. The mechanisms of these underlying cognitive models, predictive algorithms and inference engines are the key factor in determining the usefulness and effectiveness of these larger systems; these components have an outsized impact on learners and on long term changes to materials and programs.

This session will argue for the importance of open analytic models to the open education community. Key questions to be explored include: What constitutes open models, and how can these be contrasted with characteristics of closed, proprietary approaches? Why are open models necessary for systems that will increasingly guide learner instruction and progression? How do open models for learning analytics relate to the underlying philosophies of open education and science? What improvements and affordances are provided by an insistence on open analytic models? And finally, what are some examples of demonstrated dangers caused by reliance on closed models?