From algorithm outputs to classroom impact: A conversation with Marian Abu, director of teacher management at the Teaching Service Commission
This blog highlights a conversation between Marian Abu, Director of Teacher Management at the Teaching Service Commission, and Madleen Frazer, Research Associate at EdTech Hub.
Image showing a group of teachers sitting around a table and engaging in an activity. Image credit: EdTech Hub.
Like many other low- and middle-income countries, Sierra Leone faces challenges in building an effective teaching workforce. The country has one of the highest pupil-to-qualified-teacher ratios in West Africa.
To tackle this issue, the Teaching Service Commission (TSC) is taking bold, data-driven action. Earlier this year, policymakers undertook a major recruitment campaign to add over 2,000 teachers to the payroll. In doing so, they used a deployment algorithm to ensure these teachers are assigned to schools more equitably.
An algorithm is a set of instructions or rules a computer follows to solve a problem or complete a task. In teacher deployment, algorithms can be used to process large quantities of data on factors such as school location, student-to-teacher ratios, and subject demand to help deploy teachers where they are needed most.
Dive into the conversation
To support the development of the algorithm, EdTech Hub examined the factors that shape where teachers want to work in Sierra Leone. After exploring what motivates teachers to change school, we used longitudinal administrative data to understand what type of teachers move school and where they move to.
Recently, we spoke to Marian Abu (pictured left), the Director of Teacher Management at the Teaching Service Commission, to learn how the algorithm has influenced teacher deployment in Sierra Leone.
Before the algorithm, how were teacher deployment decisions made?
In 2019, we recruited 2,000 teachers and assigned them to schools randomly. This led to a lot of backlash from parliamentarians and the general public, as we could not justify our decision to deploy teachers to different schools.
So, in 2020, the TSC asked the World Bank to help us build the Teaching Service Recruitment Portal to remove the randomness from the deployment process. We used the portal to recruit 5,000 teachers based on factors like the number of trained and qualified teachers, the number of female teachers, and the need to support hard-to-reach schools.
While the portal improved the process, it did not fully address all deployment challenges. That is when we introduced the algorithm.
Can you briefly tell us how the algorithm works?
The TSC and Fab Inc. designed the algorithm to prioritise schools based on their need for teachers. For example, the algorithm looks at pupil-to-teacher ratios and learning outcomes to ensure districts and schools with the greatest need receive trained and qualified teachers.
Next, the algorithm considers the preferences of schools and the preferences of teachers. The TSC asked schools to list desired teacher attributes, such as experience, gender, and willingness to stay, and then asked teachers to specify preferred school characteristics, such as location.
If the algorithm assigned more than one teacher to the same position, it would give priority to female teachers, teachers with higher qualifications, and teachers with longer service.
Sierra Leone faces a significant gender imbalance within the teaching workforce, with one of the lowest ratios of female teachers in secondary education across West Africa. As a result, the TSC are using the algorithm to prioritise the recruitment of women.
What difference has the algorithm made to teacher deployment?
In Sierra Leone, many teachers volunteer in schools before applying for a formal teaching position and being added to the government payroll.
One of our policies states that we need to deploy new payroll teachers to schools that need them most. These schools may not be the places where teachers are volunteering.
Previously, new payroll teachers would usually stay in the schools where they had volunteered. This did not help us get more teachers to work in remote schools.
With the algorithm, we now recruit and deploy teachers to the schools with the greatest need. For example, some schools have thousands of students and only one teacher for each subject. The algorithm has been able to tell us how many teachers we should deploy to meet our target pupil-qualified-teacher ratios.
At the same time, we set a rule that teachers cannot be moved more than 5km from their previous school. So, the algorithm promotes ease for everyone.
How has the algorithm supported workforce planning?
The algorithm is really flexible, and it gives preference matching for the deployment exercise.
We will be able to improve learning outcomes, as it will help us to deploy trained and qualified subject-specific teachers to underserved areas. For example, we have a lot of chemistry teachers in the district headquarters, but not in Falaba district or Bonthe district. With the algorithm, we can prioritise getting chemistry teachers deployed to these districts.
However, the TSC will be able to speak on the success of this algorithm at the end of the academic year when we can use the Annual School Census data to see the number of teachers who stayed in their posts.
What challenges did you face when using the algorithm?
As I mentioned, there are a lot of volunteer teachers in Sierra Leone. Recruitment typically involves transitioning volunteer teachers to the payroll and deploying them to schools with the greatest need. To receive new teachers, schools need to have been approved to operate by the government.
One challenge is that some newly approved schools have not received a school identification number, so the algorithm does not recognise them. The algorithm will not deploy teachers to these schools, while teachers from these schools might be added to the payroll and deployed elsewhere.
Overall, these newly approved schools have been left worse off than they were before we introduced the algorithm. About 14% of these teachers deployed using the algorithm have been affected by this issue.
When these complaints come, the TSC issues a school identification number to these schools, and then redeploy the teachers.
Did anything surprise you about using the algorithm?
I have been surprised at how teachers have embraced the new system. The end result may not always match the preference of the teachers, but most teachers are happy to have been added to the payroll and take up their posts.
What are our next steps?
In 2025, the TSC and EdTech Hub will work with Fab Inc. and the Learning Generation Initiative to study the impact of the algorithm on teacher deployment in Sierra Leone. Here, we will focus on two main questions:
- How does the use of the teacher deployment algorithm affect the distribution of teachers?
- How does the use of the teacher deployment algorithm affect the teacher deployment decision-making process?
Afterwards, we will work with Marian and her colleagues at the Teaching Service Commission to use the results to find ways to improve the design and use of the algorithm for decision-making in Sierra Leone.
