Alelo’s Enskill platform is designed to support the development and improvement of learning content using machine learning. Enskill continually collects data from learners around the world, which make possible an innovative data-driven approach to the development of learning products.
The conventional ADDIE method for instructional design involves a sequence of discrete phases: Analyze, Design, Develop, Implement, and Evaluate. Ideally this process is iterative, where the results of the Evaluation phase inform revisions to the design. In practice this can be difficult when the designers do not have sufficient access to the target learners for evaluation studies.
By contrast, Enskill has constant access to learner data. Whenever learners interact with Enskill simulations their speech inputs are uploaded to the Enskill SimServer and saved anonymously in the cloud, which Alelo’s instructional designers use not just in iterative evaluation but throughout the design process.
For example, Alelo’s Enskill English language-learning system incorporates a model of common mistakes made by native speakers of several languages, including Spanish, Portuguese, and Thai. Designing such a model could have been challenging, but Enskill’s access to learner data made it easy. We collected example error data from an initial set of Enskill users in Chile and Thailand, and used it to create an initial model. As more learners used the system, we updated the model. The model performs better than anything we could have designed in isolation, and keeps getting better.
We are continuing to discover new ways to exploit the trove of learner data that Enskill is collecting. We collaborated with USC on a prototype authoring tool that uses machine learning to train dialogue models from example data. We are in discussions with a major online assessment provider to share data, so that we can compare the evaluations generated by Enskill against their assessments.
We are currently raising funds to continue the enhancement of the Enskill platform, for more information see www.flashfunders.com/alelo.