Creating a strong economy should be central to any country’s developmental plans. It is clear that the best choice for a country to grow its economy today is to accept the tenets of economic liberalism: keep markets free and competitive and minimize governmental interference in markets.
In developing countries, in which governments often have a stronger role to play in economic growth and development, it is essential that accurate data be used as the basic input in policymaking. The reasons for doing so are self-evident: collecting accurate data serves as the most reliable metric of the “baseline” state of a country. Changes in the data therefore reflect actual changes on the ground.
Today, we have new modes of data collection that are more reliable than those used in the past. Traditionally, surveys were seen as the main instruments of data collection. Information obtained from surveys is often accurate, but designing them is difficult and time-consuming. Thankfully today, we have a plethora of other data sources that may be more current and more reliable. These new data sources include data retrieved from the internet such as the outputs of web crawling and social media, telecommunications data, and geospatial data.
Improvements in artificial intelligence (AI) capabilities and increased computing power have given us the power to make greater sense of data. Machine learning (ML) algorithms can now process huge sets of data to make accurate predictions. The larger the data sets, and the more powerful the computers, the more accurate the predictions are. As a policymaking tool, AI (particularly ML) can be invaluable. With it, governments can set accurate targets on a vast domain of subjects.
This is critical for a country like Nepal in which much policymaking is still done in the dark, without any accurate data. For example, the education sector in Nepal has been a priority for every government since the 1950s. Despite this, the state of education in Nepal is poor. The only success that has been achieved has been in the improvement of primary school completion rates.
Private schools that have opened up since the 1990s have shown much better results. This is probably true, however, only because government-run schools perform so poorly. We have no way to make definitive conclusions without undertaking a methodological assessment.
AI may be able to provide such an assessment. Using the array of data sources described, we may be able to make more accurate predictions about where the problems may actually lie. Poverty, topography, access, and other variables may be responsible for the dismal state of education. Before, it was difficult to incorporate all these variables into economic or statistical models. Today, because of the computing power we have at our disposal, it is possible. The point is incorporating new sources of data such as satellite imagery and web analytics can give us a much richer understanding than if we had just restricted our sources of data to surveys or other traditional sources of data. We may find patterns that we could just not envisage before.
This can lead to further exploration. Other data collection methodologies, such as surveys, may be incorporated after the initial research. Research, after all, is ongoing. It is never complete. The most important contribution of AI into the current research paradigm in the development sector is that it may enable us to lay out a blueprint for taking action. We may finally have something to work on and not merely be fiddling around in the dark.
Manish Gyawali is a developmental consultant, writer and data analyst based in Nepal.