Using Artificial Intelligence for Quick Humanitarian Response

By 0
Using Artificial Intelligence for Quick Humanitarian Response

Once upon a time, there was a pandemic outbreak; a deadly virus that threatened to kill many humans and had no cure was ravaging the world. You wonder why this article starts this way. Well, it is a story that is real but is being recounted as a folktale. This story also shows the power of collaboration and innovation in an information-driven age.

So, to continue with the story, the pandemic affected all nations, rich and poor. This prompted governments to institute strict lockdown measures to limit the spread of the virus, which was airborne. Some, therefore, had to roll out financial support schemes to support their poor and needy nationals. One developing country adopted data science technology developed by a university in a developed country to tackle the problem of distributing help as cash to poor citizens affected by the lockdown imposed because of the pandemic. This was a major departure from how it was usually done in the past; previously, field teams would have been dispatched to identify and register the poor citizens for the cash distribution to be carried out. In brief, the government based the classification of poverty-prone areas derived from poverty maps based on satellite imagery.

Satellite images were used to train a data science model to identify locations in the country as poor or wealthy areas. Features such as the materials used for roofing and road types (bitumen or dirt-road) were identifiable from the satellite images available. These features were used for classifying localities or communities as rich or poor. For instance, thatch roofs indicated poor communities and shiny or colored ones (characteristic of metal roofing materials) depicted wealthy ones. The model could then be used after training to predict poverty in locations that had not been visited before. To complete the analysis, this information acquired through machine learning was then super-posed on user phone records; how often did a user buy airtime or use mobile money to make transactions, and what were the volumes of such transactions? Based on this combined analysis, financial help was sent to 30000 citizens within 2 weeks. Citizens who were in the informal economy and who needed to move each day to make daily wages to take care of their families, but whose movements were then restricted by the lockdown benefited from this. You wonder by now which country I am referring to. This does not matter; what matters is the fact that technology has been leveraged to produce a positive outcome within a short period.

The use of machine learning and artificial intelligence technology is becoming pervasive in this current era of a data-driven society. Some lessons can therefore be learned from this story for policy development on artificial intelligence research and application in Ghana and Africa at large.

  1. Academia-industry-government partnership is required:

In the story, there was a problem statement from the government; how could the poor citizens affected by the imposed lockdown be easily identified in the shortest possible time? Through its search for a solution, it identified the work done by the University of Berkeley on poverty maps creation, using artificial intelligence. In this scenario, the problem was identified on one continent and the solution was found in another. This illustrates how solutions to societal problems can be solved with tight collaboration among academia, industry, and government. In the field of artificial intelligence, research is very much required and the best centres of excellence in the domain of research are the universities. For Ghana to make the most of artificial intelligence, its universities need to fill the research gap for the identification of local solutions to local problems. In Africa, some universities have taken the lead in this initiative; the Makerere University in Kampala, Uganda, for instance, through its centre for artificial intelligence research has produced a lot of work in providing local solutions based on artificial intelligence in agriculture. The University of Cape Coast here in Ghana is also doing some great work in natural language processing (NLP). More is still needed to be done for global recognition as centres of excellence in the field of artificial intelligence research. In this regard, more government support is needed. Government support does not necessarily have to come in the form of financial support; its involvement in the problem statement and adoption of the solutions provided by the local universities go a long way to further encourage innovation in these institutions.

  1. Local collaboration is critical:

In our story, the details of the financial ramifications of the solution development are not available, but such a solution can likely come with a cost. With more local collaboration, a lot of solutions can be optimized in terms of cost to developing nations. We must bear in mind that every technology has a cost, as such importing innovative technologies, in the long run, is akin to the already existing trend of importing finished products by developing economies.

It is no doubt that the Google AI research centre opened in Accra, Ghana was in the same spirit of deepening local collaboration.

  1. Leverage on data to solve problems in novel ways:

With the plethora of data now available, the sky only is the limit in terms of the solutions that can be provided for the various societal problems facing humanity. However, this starts with clearly defining the problem to be solved and identifying how the existing data can solve such a problem. In our story, the problem was one of quick response to a situation and the solution available was the use of satellite imagery without having to deploy field teams that could take a long time and are also cost-intensive to collect the same data. To each problem, there is a different solution. Sometimes, the wrong identification or statement of the problem contributes to getting stuck in finding the data-driven solution.

In conclusion, data science and artificial intelligence technologies can be applied to different problems once there is enough data to be exploited in solution development. There are many use cases ranging from vaccination planning to humanitarian response to disasters that can be implemented using satellite imagery data to speed up implementation. A large amount of such data requires the use of machine learning or artificial intelligence techniques for effective solution development. Local collaboration ensures that the solution development is done in the same context as the data collection and, if done properly, can reduce the risk of bias in model development. Local collaboration is also beneficial in the cost management of solutions developed using artificial intelligence techniques.


Author: Yayra de Souza–Telecommunications Engineer, AI specialist (Member, Institute of ICT Professionals, Ghana)

For comments, contact / Mobile : +233543758923