Scaling AI in Education: The Importance of Feedback Loops and Interpretable Models in Driving Adoption
As industries and occupations are being transformed by AI and other emerging technologies, there is a growing need to retrain existing workers and ensure those entering the labor force have the requisite skills to succeed in the new job market. At Coursera our mission is to provide everyone, anywhere access to the high quality education that will become necessary for this. Scaling the provision of this education to millions of people will only be possible through leveraging new AI technologies and a series of best practices in their application. Through our work we have learned (1) the importance of building in feedback loops where learners on the Coursera platform can alter the models that affect their learning experience and (2) the huge benefits from interpretable models that expose the drivers of predictions to us internally and the end user. These learnings are relevant to anyone looking to drive adoption of AI in a business and scale the results of machine learning in their product.
Vinod Bakthavachalam is a data scientist working with the Content Strategy and Enterprise teams where his work has recently focused on developing ways to measure the learning outcomes from taking Coursera classes, especially in the context of company sponsored training. Prior to Coursera, he got triple Bachelors degrees from UC Berkeley in Economics, Statistics, and Molecular and Cell Biology, and his Masters degree in Statistics from Stanford.