Reflections on Education in sub-Saharan Africa: Part 2-Data for Decision Making

In working with the Directorate of Science and Technology Innovation in Sierra Leone over the summer of 2019, I found myself deep in thought about many problems and solutions in education. It was probably my most immersive experience in analyzing and thinking about the world we live in. I was typically not a person to take notes, but in doing so I found myself thinking a lot more clearly and making connections better and managed to fill an entire journal about my thoughts, questions and learnings on education at all levels. This article is a short summary of one of the thought processes I went through with the DSTI team. Admittedly, this article will be more high level and rather than mentioning concrete suggestions and actions I hope it can give you something to think about/critique and perhaps challenge you in some way.

I will describe my outlook on policy, as someone who began with little experience, and argue for why there exists a gap that can be filled with the use of quality data science. Then finally describe how we went about thinking through a particular problem related to accessing education.

What controls the education system?

Policy and Culture are the system wide controls in education. Policy determines the rules in the instruction manual and culture is how people choose to actually play the game. Culture is what happens on the ground in terms of how students, parents and teachers treat each other and perceive things in an educational setting. Ideally, policy and culture should align towards the same goal but this is often not the case. An example is when there is a rule against cheating for a public exam but the students do it anyway because it is “normal” and happens frequently anyway. Essentially, policy can differ from culture when it is not “enforceable”. Guaranteeing enforceability is a difficult task as it depends on the accountability of the enforcers, the structure of the system and the clarity of the policy, all of which are contextual. So in the drafting of policy, decision makers always have to take into account what actually happens on the ground, the reality of people’s everyday lives and not just imagine everyone will play the game the same way.

Data for Decision Making

Decision makers are constantly under scrutiny for the decisions they make. And if they make a decision, they should be able to answer why they made that decision. I do not think “it seemed like a good idea to person X” is a good enough response. Decisions that affect so many people’s lives need to be based on evidence and refer to literature or case studies of successful experiments (and not just politics which is unfortunately what usually happens). From how resources are distributed to the inclusion/exclusion of a particular topic in a curriculum, all of these decisions should be justifiable. With the field of data science blossoming in so many other fields, the same is happening in education policy and could be a great opportunity to access more information than annual reports, dependent on funding can ever provide.

Asking the Right Questions

So how can data science guide decision makers? I would argue that while the technical aspects of generating insights from data can be difficult, understanding the insights is probably more difficult and a lot more important. What you want to produce as a data scientist are numbers, plots, maps, animations or other types of visualizations that ultimately aim to answer a question. The part where a lot more careful thought has to be put in is in making sure the correct questions are being asked and if they are being answered using the correct method. This is often where I get stumped if I try to think about it more deeply as opposed to just making random plots with a dataset.

In trying to think about solutions to problems, I often want to imagine myself as the student and ask how a decision will change anything about my experience. How can I possibly understand and empathize with the incredibly diverse difficulties students might find themselves in? The answer is that I cannot fully understand, nobody can. That means that in developing insights and models, I must constantly be looking for feedback, tweaking, improving and creating new ways to look at things via consultation with others. The goal is to be able to confidently be able to respond to all the questions that begin with “But what if …”. In many cases, you can do so by understanding a variety of different case studies and learning from the past, something which all policy experts ideally strive to do. Alternatively, you can perform experiments through which you can test policies and measure outcomes.

So then how can access to education be measured? There is a wealth of literature that exists for this, my favorite is the Center on Great Teachers and Leaders website. There are of course many variables that come into play, some factors such as the distance a student has to travel to school, the direct costs (tuition) and indirect costs (uniforms) etc. are more intuitive. Some less obvious ones include the student:teacher ratio, cultural norms that might exclude for example girls from going to school as easily, nutrition and many more. In an attempt to improve literacy in Sierra Leone, where 60% of adults cannot read or write, the Government declared on Primary school education free and dedicated over $1.5 million dollars to education. This should go a long way to reducing the barrier to access caused by direct cost but it is certainly not the complete picture.

Given that primary school is essentially free now in Sierra Leone, the next factor we wanted to consider was if children had difficulty physically accessing schools. This translated to two (in addition to other) questions, “How far away do people live from the nearest school?” and “How far do they have to travel to get there?”. This is where the data science comes in. We managed to get census data on the general locations where people lived as well as the locations of all the schools in Sierra Leone. Using this data and template code by Adam Rahman, from UNICEF for performing efficient nearest neighbor searches for geospatial data, we then could estimate the average distance each person would have to travel to a particular kind of school within each administrative region. We felt this was a good first step in measuring the difficulty different communities of children face to get to schools but it is only a first step.

The results were combined with other interesting metrics such as the estimated children not in schools to find correlations, some interesting, some not so much. But something I like to emphasize is that all of the results must be assessed thoroughly before coming to conclusions. This involves investigating in the field, using different datasets and most importantly asking even more questions of the right kind, some scientific, some more human in nature. What is your estimated error? What is the condition of the roads used in routing? etc. The results then need to be visualized and well explained so that the policy makers understand them and are not misled. This is where again I value the use of good research questions.

Conclusion

My aim was not to bore anyone with the details of the process we went through but I would really like to leave a challenge. In this digital age, there exists so many datasets available and so many questions that could be answered by them. The barriers to academic publishing are ridiculously high, the exposure and quality of a typical project from a school or personal project is too low and can often not be very relevant. Non-profit and company reports are kept on dispersed websites with little traffic. Can we find a middle ground where students, aspiring data scientists and the open source community in general can guide and review each other, much like Medium? Can such a centralized platform be used to answer key questions that can solve social issues in a way that benefits everyone? This is one of the many questions I jotted down in my notebook.

If you would like to comment on this article or this series in general, feel free to comment below. Next time I will talk about the disconnects I see between universities and societal needs and make some suggestions, much like in part 1.