By Bapi Chatterjee
Designing a curriculum for an ever-changing academic landscape is challenging. In addition to teaching the foundational concepts, a well-structured curriculum
must nourish the critical thinking and problem-solving skills of students, and, should adapt to changing needs and opportunities in the market. Emotions, personal experience, and anecdotal evidence of a faculty member inordinately influence the design of a curriculum. A better approach is the application of data and statistics. With statistics coming into the picture, its more glorified avatar, “Machine Learning,” finds a natural application in this context. Here we derive insights from recent research in this domain.
Discussion forums like Reddit and Stack Exchange are extremely popular among computer science students. Karbasian and Johri [1] used a machine learning approach, namely latent Dirichlet Allocation, to suggest essential topics and valuable examples for professional development to incorporate into a data science curriculum using data from these two portals. However, their study also suggests that often emerging topics, for example, ethics in AI, can be missed because of bias in the collected data.
Ball and co-authors [2] apply a machine learning method called logistic regression to identify the features from students’ transcripts, such as age, gender, a first math course, international status, etc., that best lead to graduation. Such an exercise helps identify the group of students that requires recommending different courses to credit in a specific order. Personalizing the curriculum is in demand in modern higher education. Applying machine learning to enhance students’ preparedness, thereby improving their critical thinking and problem-solving skills, can be an extremely efficient approach to designing a personalized curriculum.
Rawatlal [3] uses another machine learning method, namely decision trees, to propose the prerequisite structure of a curriculum based on historical records to determine the effective progression routes for students. Artificial neural networks, arguably the most popular machine learning methods at present, were applied by Somasundaram and co-authors to define curriculum with the aim of bridging the gap between the current state of a program and its expected outcome. They exemplify their proposition for identifying the curriculum required to meet the job roles in the area of the Internet of Things (IoT).
Malik and Gangopadhyay [4] review the applications of machine learning techniques in different phases of the end-to-end education process —from planning and scheduling to knowledge delivery and assessment. Theirs is an excellent resource for understanding the evolving role of machine learning approaches in curriculum design over the last two decades.
Indeed, an unpretending data-driven approach prominently employing Machine Learning is the need of the hour for modern curriculum design.
The author is assistant professor (CSE) at IIIT-Delhi. Views are personal.