Roles for computing in social change, Rediet Abebe, Solon Barocas, Jon Kleinberg, Karen Levy, Manish Raghavan, David G. Robinson, FaCCT, 20 https://arxiv.org/abs/1912.04883
Without overclaiming computational research’s ability of directly solving social problems, this paper lists out 4 valuable roles that computation can play in the service of social change. Here I’ll talk about 2 roles that I like.
Computing can help to diagnose how social problems emerge in technology. That is, it plays an important part in interrogating itself. The paper lists several examples, for instance, how researchers utilize computation methods to identify gender biases in word embeddings or racial biases in facial analysis systems. This reminds me of the InfoSeeking project I am currently working on. Our topic was to identify biases in images returned by search engines. We compare different combinations of search keywords, languages, and search engines to try to demonstrate the difference and possible biases in the groups of images returned. Here, the search engine itself is a “computing method”, but the difference it reveals roots more deeply in social problems. For example, when we search “successful person” on Google, most of the returned images are about males. This identifies possible gender bias that may come from the injustice in recruiting processes. In this way, computational approaches provide support or evidence in measuring social problems in technology more precisely. However, the paper also stresses something important: we should not confuse diagnosis with treatment. Although the computational diagnostic studies might drive real social changes, e.g. the findings of the facial analysis biases actually stimulated improvements in Microsoft and IBM’s facial analysis technologies, computational system itself is usually an insufficient remedy of the problems it reveals. There are still a great number of problems in society that we know exist but have no good ways to address.
Computational approaches can bring long-lasting social problems to the public eye once again. The paper states that “technology critique often captures public attention”, for example, “many people would not pick up a book about poverty policy in general — but are game to read a critique of the algorithms used to administer it”. However, the reason for this statement is not clearly stated. Here I provide a possible explanation: computational researches are highly reliable. As math and science are definite, the computing results can be a strong support to demonstrate the problems (this is similar to the computing role above). Rather than just stating that “poverty is a serious social problem that we need to address”, providing specific evidence like “researchers show that there are 70% of people in country X who live in poverty” would make the problem salient and thus more likely to catch public eye and leverage resources that might not accrue otherwise. One thing to notice is that computing only acts as a synecdoche and offers us a lens that peeks into part of the problem. Social problems are complex in nature, which makes it difficult to focus on them as a large whole. A common method is to cut them into small pieces and address each piece separately. However, it is significant to not be constrained to the technological aspects of a problem. Excessive focus on computing lens would mask the root problems in society as well as the non-technological ways of addressing them, which ultimately “circumscribes the issues that can be brought to the table”.