Could Artificial Intelligence grade your next essay?

While professors can teach a nearly infinite number of students at the same time, their ability to grade their students’ work naturally limits class sizes. Multiple-choice quizzes can be graded rapidly, essays on the other hand take up more time. A growing number of Start-ups and universities are now looking to solve this problem through robo-grading that is powered by Artificial Intelligence. First advancements in the field of automated grading of essays were already undertaken in the 1960s by Ellis Page, who was a researcher at the University of Connecticut. At the time however, computers were not powerful enough to get the job done. Fast forward five decades later in the year 2012, the William and Flora Hewlett Foundation initiated a competition to find the best automated grading startup. The robo-graders were given a number of essays that were graded by humans. The startup that got the closest to the grade of the teachers would win a price of $60,000. 154 teams participated and the winning team was in 81% agreement with the human graders. Since then, computation power has increased significantly and AI has advanced from computer labs into economically viable solutions.

In 2016, two researchers from Stanford University claim that they could have achieved a coincident of 94.5% in the robo grading competition. Their system works as follows: A certain number of essays that were graded by humans will serve as a basis for the algorithm to learn. From this point on, the algorithm can take over and grade the remaining essays. The AI can understand structure, clarity and reasoning of the essay. However, if the sample is not large enough, the algorithm will be imprecise and might even draw wrong conclusions.

In my opinion, essays are the most meaningful assessments of a student’s knowledge, yet I have experienced that they are being replaced by short answer exams and multiple choice questions in many universities. I am hopeful that faster and advancements in AI will allow robo grading to make essays the primary form of assessment again.