An assignment index for Professor Frazier's DATA 100 class
There are three big components that most people consider when defining a nebulous concept like computer science and programming: hardware, software, and people. After all, it is people that ultimately created the technology allowing us to deposit money to our bank accounts 20 miles away from the nearest ATM, instantaneously locate the nearest gas station when we’re driving low on fuel in the middle of unfamiliar territory, and see the faces of our loved ones appear on a screen as if they’re right beside us, even if they are video-calling from a different country. Exploring the “harder” side of this realm of science leads us to appreciate algorithms and code that can be applied to a myriad of problems to create automated emergency alert systems that know how to respond to floods before they happen, or analyze the correlations between benchmarks of development and the number of young people that comprise a region’s population.
But when it comes to one of the most applicable offshoots of this field – data science – it is all too easy to forget what the products of experimentation are meant to produce. Especially in an era when the top 1% of the wealthiest in the world own more than a third of all global resources, it is vital to adapt the tools we have to benefit as many of the needy as possible, as quickly as possible. Whether a population’s defined need stems from gender-based inequity, racial or ethnic bias, fiscal inequality, or zip code-based discrimination, there are solutions to be found in big data. However, code and algorithm developers too often dismiss the biggest problems facing our modern society as untouchable, preferring perhaps instinctively to focus on the theoretical; the virtual, instead. It is in this context that the primary plea of Nature contributor Joshua Blumenstock rings clearest: “Don’t forget people in the use of big data for development.”
Blumenstock’s biggest gripe with “big data” can be summarized by its apparent inapplicability, or, more specifically, its unconscious inapplicability. With the majority of digital aid organizations, social media companies, and data researchers isolated in pockets of affluence throughout the globe, it is understandable – but not acceptable – that most of the products of these groups wind up failing to carry out their intended purposes when they are put to the test in the real world. For a prime example of this kind of unanticipated foundering, look no further than the failed “one laptop per child” project. Many big data solutions, while promising in prospect, ultimately crumble under the weight of reality; the most efficient algorithms cannot afford to be detail-oriented, or risk succumbing to the crushing weight of superfluous data. But seeing the big picture at the expense of the small brushstrokes that comprise it means overlooking people and issues that are already poorly represented. Populations that survive with little to no social resources (e.g., literacy, internet connectivity, etc.) will only become less visible to digital data-based aid programs, which rely on pings from cell phone calls, online interactions, text messages, and GPS tracking to identify physical areas of need and send resources to the people who inhabit them.
By elucidating the deadly impacts of both unpredictability and bias in algorithmic solutions to development problems, Blumenstock’s clamor for the refinement of his field is given undeniable strength. It should be clear to any reader now that big data needs both validation and regulation to successfully catalyze the process of human development, which is exactly what Blumenstock suggests. Along with attributing a better sense of context to algorithms to make them “FAT” (fair, accountable, and transparent), Blumenstock also proposes that deepening collaboration and partnerships between the private and public sectors; government and NGOs; and political and data scientists could help bridge the widening rift between big data solutions and real social change. Collaboration between existing and novel datasets and data collection methods, such as the joint phone- and survey-based data-gathering system currently being utilized to enumerate the positive impacts of the World Food Programme in Haiti, is another way big data implementation could be altered to better benefit humans in place of pure technological progression.
According to Blumenstock’s article, “95% percent of the global population has mobile-phone coverage,” providing motivated analysts, programmers, and social activists with a data-based tool that has the potential to span racial, ethnic, socioeconomic, gender, and geographic boundaries in its ability to collect information. This does not mean that technology can do all of the work – algorithms are inherently flawed and will always carry the biases of those who write them. Lacking transparency and understanding between the inextricably-linked creators, consumers, and regulators of data-based non-profit aid programs causes problems down the road for those seeking to implement lasting change. What’s more, the dangerous allure of cheap digital data belies its inaccuracies, glossing over discrepancies that exist between localities when blanketing correlations are publicized as one-size-fits-all solutions to global problems. Clearly, there is much to be improved upon when it comes to actually applying big data to the real world, especially if genuine, equitable results are hoped to be achieved. This is all to say that there is a simple answer, albeit one that entails many complex processes to be carried out in the future. If developers hope to use big data for the betterment of mankind, they must live up to their title – and focus on human development, not just human-borne data.
Blumenstock, Joshua. “Don’t forget people in the use of big data for development.” Nature, 10 September 2018, https://www.nature.com/articles/d41586-018-06215-5.