Combatting poverty has been a hallmark of good development strategies. Ever since the birth of the Sustainable Development Goals and the 2030 Agenda in 2015, researchers, academics, and now corporations including IBM are looking for ways to help boost economies and sectors within them that are struggling. Technology has been a big part of that, and now with Machine Learning (ML), hopefuls are looking to combat unnecessary suffering and lift people up.
But practitioners need to be careful. Especially on the corporate side of things, whole organizations have praised technology and the private sector for what they can import to developing economies, without considering the adverse effects and long term evaluations that will need to be conducted by researchers, not technologists, on how successful certain tools will or can be.
When using ML/AI to combat poverty, it is one thing to understand using satellite imagery where rough neighborhoods are, it is another to determine the efficacy of a technology that uses ML in those neighborhoods. The difference needs to be understood and analyzed. And people will need to be trained on how to use these new technologies.
Educational Outcomes and ML
Cio.com delivers the latest tech news, analysis, how-to, blogs, and video for IT professionals. One contributor claimed in 2019, when discussing the power of AI to combat poverty:
“As long as access to computers and the internet are available which may include satellite communication networks, AI-teachers can provide students education based on a controlled syllabus.”
While it may be true that such technologies will reduce labor costs, and allow greater access to more data for students, thus leveling the playing field a little bit, there are still many other considerations that need to be acknowledged.
While employing ML in economically underserved communities in the U.S. or abroad is important, there’s too many other variables that come to mind that affect educational outcomes. In developing economies, it might not matter if there is Internet available in the classroom if students don’t have decent roads to get there. It also doesn’t matter if good technologies exist in the classroom if students are forced to stay home to help with household tasks. Inequalities exist in education because of the way life goes on outside of the classroom. This perspective should be considered for technology giants trying to understand variables in economic development, namely education.
While ML solutions in education have to account for multiple variables, there is also caution that needs to be taken with regard to the agricultural sector. Recently, thanks to IBM and Carnegie Mellon University’s project FarmView, researchers are using drones to improve crop management techniques. They claim on their website that AI is tackling the world’s food crisis.
The idea behind the project, from a farm management and crop rotation perspective, is that drones can identify drought prone regions and researchers can then plant drought resistant seedlings in those areas, maximizing crop yield per hectare.
FarmView does other things too. They are experimenting with a robotics/AI system that detects the size, number and quality of different fruits grown on an experimental plot of land.
Stephen Nuske, a systems scientist in the Robotics Institute and influencer on the project, wants to make these robotics systems affordable to small and medium sized growers, although doesn’t specify how he is going to get these systems off the ground and into the hands of the most vulnerable farmers in far away places.
While Machine Learning in this sense has a lot of practical potential, it is again, like the education predicament, where confounding variables (such as droughts or seasonal flooding, or the breaking of equipment), that causes reason for concern when it comes to ML in this context. Will existing farmers be able to train others once the technology is passed down? These things should be considered in part of the design process. That’s why the work of technologists and social science researchers are different, but equally important.