Disrupting Education: Which Algorithms Make the Grade?
By Samantha Bansil, Matthieu Kovel-Lazzarini, Gloriana Lang-Clachar, Alexandra Not and Aizhan Shorman
Algorithms are everywhere. They influence many of our choices, from where we eat to who we date and what we binge on Netflix. Only recently have we begun to explore the use of algorithms to improve the decision-making process in a variety of sectors. This can have an enormous impact in the realm of education, where algorithms can help both teachers and policymakers better serve the needs of their students. In this post, we will explore how school systems around the world have started to implement algorithms. Perhaps it’s time to sit back, relax, and let computers teach us a thing or two.
Student Placement in U.S. School Districts
The United States has a somber history of racial discrimination. Although segregation has been outlawed for several decades, its legacy can still be felt in the public school system. To this very day, the vast majority of American students still live in racially concentrated school districts. Various schools have struggled to bridge the divide through numerous policies, such as San Francisco’s attempt to racially integrate schools in the 1970s by bussing children from one neighborhood to another.
Today, more districts are trying to establish a more equitable way to enroll students through the lottery system. The My School DC system is a prime example of this. Based on algorithm designed by economist and Nobel prize recipient Al Roth, My School DC aims to match students to open slots in public and charter schools through specific criteria (such as the family’s preference and distance from the school).
It is a unified enrollment system, meaning that the family only needs to submit one online application for consideration by all participating schools. Before, schools had separate applications and deadlines — a time-consuming and costly process for families with limited resources to navigate. Proponents of My School DC believe that it helps level the playing field, giving poorer students access to better quality schools. Theoretically, the algorithm is more transparent and efficient, as well as a better way to account for student preferences.
However, issues remain with regard to transportation time: although the algorithm claims to take into account the distance from the student’s residence, it can still place students in schools that involve very long commutes. The algorithm furthermore does not resolve all inefficiencies: even after two rounds of applications, 42 percent of all participants were still not matched to a school in 2014.
Ultimately, one algorithm cannot undo the structural effects of segregation, nor can it fully bridge the digital divide between the rich and the poor. There is still a jarring difference in the amount of quality resources available to wealthier schools compared to less-fortunate ones. My School DC helps families and individuals access better quality education. But at a broader level, it does not target the root of the problem or assist underfunded and under-performing schools.
Our grade for My School DC? B-
Teacher Assignment in Italy
According to the Programme for International Student Assessment (PISA), Italy lags behind its OECD counterparts. Education performance has decreased in recent years due to various factors, such as lack of funding and an increasing number of students per class. The government attempted to reform the system with the Buona Scuola project, which was developed with a variety of goals in mind (including the hiring of more teachers and adoption of digital innovation and pedagogy), as well as a highly controversial algorithm.
The algorithm, designed through a collaboration between Finmeccanica and HP Enterprise Services Italy, emerged out of a public competition. Its primary aim was to match the best teachers with the schools with the highest need over the 2016–2017 school year. Although the inner workings of the algorithm are unknown, it was supposed to take into consideration the teachers’ location preferences as well as their evaluation score (which served as a quality indicator of their teaching abilities).
When put into practice, however, the algorithm did not assign teachers as planned. Instead of placing higher importance on the evaluation score, the algorithm assigned teachers mainly according to their first destination preference. This was due to poor programming, coding errors, and other technical problems. Many teachers filed complaints, eventually leading to a court case. As of 2019, the algorithm is no longer in use.
Apart from the Buona Scuola reform being vastly unpopular among teachers and unions, many claimed that it exacerbated the existing inequalities within the system. Opponents of the algorithm argued that it unfairly promoted private schools at the expense of underfunded ones, and that it excluded substitute teachers. Other critics further claim that the algorithm stripped teachers of their agency and that it lacked accountability toward families and students.
Our grade for Buona Scuola? F
Adaptive Learning in Uruguay
Governments are trying to harness technology’s potential to facilitate an entirely new, adaptive way of learning. This is the case in Uruguay, which launched its transformative Plan Ceibal initiative in 2007. Aiming to bridge the digital divide within the country, the initiative has succeeded in increasing the use of ICT in schools, especially through its “one laptop per child” policy.
As part of the Plan Ceibal program, Uruguay has developed an online, adaptive learning solution called the “Mathematics Adaptive Platform” (PAM being its Spanish acronym). PAM offers teachers tools to both work with and evaluate students, as well as allows students to learn at their own pace. Through its algorithm, the platform can identify areas of improvement and thereafter adapt its content to target each individual student’s needs.
Plan Ceibal has a positive impact overall. Studies show increased motivation for both students and teachers, and most parents approve of the program. Plan Ceibal’s egalitarian implementation further contributes to its popularity: the program first rolled out in rural and poor communities, which stands in stark contrast to how educational innovations are usually limited to wealthier areas at first.
Critics of Plan Ceibal mainly question its effectiveness and efficiency. Towards the beginning of the program of the 2009, the Economist pointed out that half the computers in one Montevideo school were broken in the first year of Plan Ceibal. Computers also initially had software installed in English, which the students could not understand. There is also a lack of infrastructure in some regions that significantly limits connectivity — a gap that Plan Ceibal currently strives to address.
Our grade for Plan Ceibal? A
Lessons Learned
Above all else, artificial intelligence will never be the cure-all for any education system. In order to have a positive, wide-reaching impact, algorithms need be transparent, properly implemented, and adapted to each context. Inequalities will continue to persist — but with the right policies (and the right technology), we can move our schools in the right direction.
