Online Learning’s Grading Challenge
In the last few months, we have seen the acceleration of online learning throughout the globe. There is no doubt that this trend is here to stay and will ultimately transform the educational space. How deep the transformation will be will depend on the equilibrium we reach on the other side of this pandemic. A critical area that we currently face in online learning is grading at scale.
In the online learning world, we are seeing enrollment figures increase exponentially. Individuals from around the world are eager to learn new skills so they can stay competitive in this ever changing global landscape. In order to truly assess if a student understands the course’s concepts and materials, there must be a proper grading infrastructure in place. Given the high enrollment figures, institutions do not have the bandwidth nor the resources to grade every student’s assignment. Most institutions have deferred to multiple choice questions (MCQs) and peer grading when it comes to assessments at scale. While there is value in MCQs and peer assessment, these mechanisms can only assess so much. For many subjects, open-response-assessments (ORAs) and essays are essential to deeply test a student’s understanding. If we are to ensure that online learning delivers on its promise of true learning, then we need to find innovative solutions to deal with the grading challenge.
Artificial intelligence (AI) can play a pivotal role in alleviating the grading process. While I think humans will always be an important part of the process and give comprehensive feedback, AI can take the burden off of instructors and assess key aspects of an assignment in an efficient manner. There are existing technologies that I have seen, but they have fallen short in two key ways. First, they require a lot of training. Machine-learning approaches need tremendous amounts of data, which is not always feasible for certain courses; more importantly, institutions cannot always assure that they can provide diverse and broad data sets that are not susceptible to bias. Secondly, I have not seen the transparency we need around these AI solutions. While black box technology is acceptable in certain use cases, it will not work when it comes to grading. If students, instructors, and institutions cannot understand the reasoning behind a machine’s grading, then they will not trust it. In education, the stakes are too high, and we can’t settle for black box technologies.
This is a critical moment for us. Online learning has offered the promise of democratizing education and ensuring that everyone can access it, no matter where they live, or what their socioeconomic background is. Our bar must not be the number of students enrolling in our courses; instead, it must be the quality of education of which accurate assessment at scale is a key component, and ensuring that students truly grasp the concepts and material that they are being taught. Despite the current challenges in AI and grading, I’m optimistic that we will see more solutions that are data-light and embrace transparency. If we can solve online learning’s grading challenge, the transformation that education will undergo can take learning to a new level.
– TC Haldi, Senior Director of MIT xPRO
About the Author:
TC Haldi oversees MIT xPRO for the corporate and professional market. She and her team partner with faculty and groups across campus to design and deliver online programs that transform individuals’ skills and drive meaningful business results. Before working at Open Learning, TC was the Director of Content, Learning Design, and User Experience in the Corporate Learning unit at Harvard Business Publishing. She holds a BA from Harvard College.