Under normal circumstances, developing a school schedule can be a cumbersome task for school districts. Many districts spend as many as six months developing schedules for the upcoming school year due to rapidly-evolving state requirements. But the onset of COVID-19 brought on new complexity to this already-difficult problem: Schools now had to quickly figure out a new scheduling strategy that complied with social distancing guidelines and minimized close student-to-student contact with as few class enrollment changes as possible.
When it comes to scheduling, humans can find strategies that work, but the process is often painstakingly slow, and the answers are rarely optimal when managing complex constraints. School environments became far more complex in 2020. Each district scrambled its way to completing the 2019–2020 school year through either in-person instruction, at-home instruction, or a combination of the two.
In the summer weeks that followed, there was only a short amount of time to come up with a schedule that would meet the needs of state requirements, students, parents, teachers and COVID-19 protocols from the Centers for Disease Control (CDC). While school professionals could put together a solution independently, finding the optimal solution—and finding it quickly—was a near impossibility due to the sheer number of factors to consider and analyze.
Algorithms can empower decision makers to make optimal decisions in extraordinarily-complex circumstances by providing a rigorous foundation for making comparisons between alternatives. Knowing this, one of Indiana’s top school districts reached out to Vertex Intelligence, a data science company, to request assistance with their school scheduling dilemma.
Vertex Intelligence offers optimization models that have been proven to save thousands of hours of trial and error while providing a statistically-sound basis for critical decision-making. For the local school district, Vertex Intelligence was able to create a model that addressed the incredibly-unique circumstances the pandemic caused.
It is important to note that the goal was not to predict how many students would get COVID-19, or how quickly it would spread throughout the student population. The goal was to identify which strategies would minimize connectivity between students.
At the time, there were several ideas about how to achieve social distancing in schools, like: alternating day cohorts, home-rooming, hybrid virtual/in-person classes, and more. With the analytical tools to help remove the guesswork out of which scheduling strategies would be most effective, Vertex Intelligence went to work looking for a way to strategically reduce connections between individuals in order to minimize the risk of transmission throughout school populations.
Unlike digital social networks, the social interactions between students throughout the school day are largely prescribed by their class schedules. Given the company’s expertise in applied network analysis, Vertex could predict the relative impact of different scheduling strategies based on the structure of each school’s social interaction network.
By leveraging direct data integrations to the school calendar, Vertex Intelligence developed a system to mine district-wide class schedules to create a highly-accurate reconstruction of probable extended close contact networks of teachers and students within the school day. That data was then integrated into Vertex’s proprietary artificial intelligence (AI) based scheduling optimization algorithm to search for class schedules that could adhere to various social distancing protocols while making the fewest number of modifications to planned course enrollments. Vertex Intelligence was then able to quantify how much impact each strategy would likely have on overall connectivity within the student population.
Vertex Intelligence used AI to analyze three scheduling strategies:
- The existing, standard in-person schedule
- A 50 percent cohorting strategy (where students come to school on alternating days but preserve their normal class schedules)
- An “extended block” scheduling strategy that was identified by the Vertex Intelligence optimization algorithm (where students have classes broken up into two blocks, each with three classes. Schedules toggle between the two extended blocks, allowing students to be in school in-person each day, but limiting the number of students they were around each week)
Using a network analytical approach, it was determined keeping the standard schedule would result in the most close contact between students by far, the 50 percent cohorting strategy would result in a 33 percent reduction in close contacts between students, while the extended block scheduling strategy would result in a nearly 50 percent expected reduction in connectivity.
Additionally, Vertex Intelligence extended its network approach to look at the overlap in students between classes to identify which individual classes would contribute the most connectivity risk by bringing together students who would not otherwise be likely to share classes.
Music classes and physical education classes were deemed the greatest risk within grade levels. However, Vertex Intelligence also noticed some courses that combined grade levels—such as foreign languages and math courses—were the most probable venues for so-called super-spreader events.
As a result of the data from Vertex Intelligence, the school district opted to implement the recommended extended block schedule alongside some remote learning courses. Thanks to Vertex Intelligence’s ability to run multiple, complete scheduling scenarios in a matter of days, instead of months of manual efforts, the school district was able to make a decision based on data, and inform parents of the new schedule sooner than many other districts.
While many school districts are still unclear on what school will look like for the 2021–2022 school year, what is clear is that AI can help make those difficult decisions a lot easier.
How to leverage AI for planning enterprise data warehouse migrations
Competitive Intelligence and Performance Optimization Models Provide Smarter Insights for Manufacturer Selling Products Online
How manufacturers can improve online retail sales performance with better competitive intelligence