In his Nature article, “Don’t Forget People in the Use of Big Data for Development,” Joshua Blumenstock argues that humility and the consideration of people in data science is imperative in order for it to benefit the real world. In order to showcase the potential of big data in tackling global issues, Blumenstock describes how data from mobile phones and satellites can “be used to match resources to people living in poverty (3)” as well as “track the effects of a natural disaster on individuals minute-by-minute (4),” which shows that big data is invaluable for the purpose of gathering and evaluating information. However, Blumenstock immediately follows by describing the shortcomings of big data: proposed solutions based on what big data shows are short-sighted, such as how digital loans in Kenya have unintentionally resulted in numerous “poverty cycles and debt traps (5);” big data can lack validity, since its patterns cannot be generalized, paradigms are constantly changing and people naturally game the system (5, 6, 7); bias in algorithms is prevalent, particularly undercoverage as a result of “vast segments of the population in developing countries (7)” not having access to technology; and finally, as a result of few “checks and balances (8),” there is much corruption and a focus on personal gain by the owners of the data. By describing the many downsides of big data, Blumenstock shows how it is insufficient to rely on it for designing practical solutions, supporting his idea of the importance of recognizing the shortcomings of data science and considering humanitarian ramifications. Having established the importance of humanitarianism and humility in data science, Blumenstock provides some steps in the right direction to mitigate the shortcomings: collaborating the use of big data and “existing methods (8);” taking “local context into account (9)” in addition to using algorithms; increasing collaboration between data scientists and organizations (9) as well as “technical capacity locally (10).” These “ways forward” embody Blumenstock’s idea of “a humbler data science (10),” which if executed properly could “transform international development (10)” as opposed to cause large-scale harm.
Anna Raymond’s response, “Good intent is not enough in data science when dealing with the problems which determine people’s experiences,” is illustrated by Blumenstock’s example of digital loans in Kenya. Although the lenders had the intention of expanding the ability to receive credit to people who otherwise couldn’t, the digital loans generated a swath of new problems like “poverty cycles and debt traps (5),” which shows that “good intent” in data science is short-sighted and insufficient to actually help people. Regarding Nira Nair’s response that “Transparency is the underlying issue to many of these problems, so an increase in this on both ends (data based issues & human based issues) could lead to better results,” it is not always the case that transparency can lead to better results. Blumenstock exemplifies this by describing how people “are inevitably incentivized to game the system (7)” when aware that their data is being collected, which shows that in some cases, transparency decreases the accuracy of collected data. Finally, Kayla Seggelke’s response that “In lieu of such drastic potential for promoting applications yet demoralizing hindrances, the balancing act can become difficult,” is addressed by Blumenstock’s description of combining big data with the use of “conventional data sets (8).” Blumenstock addresses how big data contains bias as a result of undercoverage (7), but by accounting for the undercovered population via “existing methods (8)” like conventional surveys, it is possible to balance the large-scale operation of big data with the slower, more traditional methods of collecting data in order to obtain a more complete picture of complex global issues.