AdmitHub Remember how much we talked about college when we were in high school? Teachers emphasized it; colleges sent ambassadors to talk to our classes; we even had a counselor whose entire job it was to help us make our college decisions. These efforts should be applauded. There’s no question that a college degree equates to higher incomes, on average, among American workers. Some studies put the income difference as high as 74% between high school and college educated workers. There’s hardly anything more valuable a high school educator can do than encourage a student to attend college. However, college doesn’t begin the day after high school graduation. If it did, it would be an easy handoff from the supportive network of high school teachers and counselors to the various orientation programs offered by colleges. In reality, there is a two month gap of summer vacation between a student’s last interaction with their high school support network, and their first interaction with university staff. In that time period, roughly 14% of all students planning to attend college will change their minds the summer before fall enrollment. This phenomenon is called “summer melt,” and is a major source of concern among educators, particularly in how it disproportionately affects low income students. Up until now, proposed solutions to the problem have been too resource intensive for universities to seriously pursue, but now, thanks to artificial intelligence, there appears to be a solution. Why students quit school before their first day In the movies, there’s always some emotional conflict that keeps a promising child from attending school. They have to stay home and take care of a sick relative, or they can’t bear to leave their high school sweetheart. While those tropes certainly happen in real life, in the majority of cases, summer melt has a much less dramatic cause: The university enrollment process is complicated and difficult for 18-year-olds to navigate alone. According to a study done through Harvard University, lower income students face an even more difficult enrollment process, as their enrollment is typically complicated by programs like financial aid, and they are less likely to have access to supportive resources. For example, among Boston Public Schools (a district that serves low income students), roughly 20% of college-intending high school graduates did not enroll in college the following fall. From a distance, the solution seems obvious. Without the resources they had in high school, young people are less likely to complete the college enrollment process, so why don’t universities engage graduates earlier in the summer? Well, there’s an issue with that solution. Universities have the answer, just not the staff It’s been documented for years now that summer melt can be decreased by early intervention and engagement with students by the university. As a result, many universities have made an effort to engage prospective students with peer mentors and info sessions. The problem is, this solution takes a lot of resources. Even with more passive methods of communication, like text message or email, a small team of university employees can’t possibly walk thousands of incoming students through the various hurdles in the enrollment process. Increasing the size of these support teams is one way to solve the problem, but there are obvious problems. At a certain size, these teams are just not financially feasible to the university, meaning that the university can only afford to field a team too small to service all of their students. Colleges have had to make due with this incomplete but better-than-nothing solution for years now, but finally, thanks to developments in artificial intelligence, there may be a more complete solution available. How Artificial Intelligence solved summer melt Georgia State University has been dealing with summer melt for a long time. With roughly 32,000 undergraduate and graduate students, they bring in a large number of freshmen every year, and therefore lose quite a few students to summer melt yearly. Administrators at GSU contracted my company, AdmitHub, to combat summer melt. Our strategy was simple. We developed an AI-powered chatbot called Pounce, and taught it how to engage students using over 80 pages of FAQs and process documentation from GSU administrators to guide them through the summer months. Pounce was able to send over 50,000 messages to 3,100 students, helping them navigate all aspects of the enrollment process. Less than 1% of the interactions needed to be escalated to human administrators, and 94% of students recommended Pounce be given to the next incoming class. As a result, GSU had the best enrollment results in school history. Summer melt was lowered by 21.4%, increasing overall enrollment by 3.9%. This is just one case, but if these effects can be replicated in universities across America, the results could be massive. For context, as of 2012, there were 21 million students in higher education in America. The experiment proves that universities have all the data necessary to engage students and help them through the enrollment process, they just lack the personnel to actually share it with prospective students. With artificial intelligence, those personnel costs are greatly diminished, until every student has access to the information and resources they need to get to college.
- This article was originally published on Tech Talks. Read the original article here.
- Andrew is the co-founder of AdmitHub, an AI startup that helps colleges engage students with chatbots. Previous to AdmitHub, he founded Signet Education, which employs more than 125 educators, and KarmaNotes.org, a non-profit knowledge-sharing website popular at Harvard and MIT.