GEOG 323 Reflection 5: Humanitarian GIS
Reading Crawford and Finn's (2015) article "The limits of crisis data: analytical and ethical challenges of using social and mobile data to understand disasters" opened my eyes to some of the limitations and concerns related to the use of volunteered geographic information (VGI) such as that obtained from social media. Given the ostensible ubiquity of social media in our day in age, it's easy to conceptualize social media data as a relatively accurate portrait of global human opinion at any given time. However, as Crawford and Finn argue, the case is not that simple. The authors point out several sources of uncertainty arising in the use of VGI for research on disasters and humanitarian crises.
First, the very nature of natural disasters means that power outages and disruptions in wifi and data access are common. Thus, rates of social media posting are likely to decrease immediately following a disaster event (Crawford and Finn 2015, p. 492). Another source of uncertainty is that while "[s]ocial media datasets depict a specific time period, typically defined by the spike in Twitter messages or the use of particular hashtags," the periods of aftermath and recovery are also key pieces of how disasters manifest and are experienced (Crawford and Finn 2015, p. 493). Thus, we only get part of the story: "The focus on social media, and the hashtag in particular, can miss the full scope of an event, as well as the long-term reasons for and implications of the disaster that are highlighted by critical disaster studies" (Crawford and Finn 2015, p. 493).
Furthermore, any given source of VGI is not representative of a place's population as a whole but is naturally skewed toward particular demographics (Crawford and Finn, pp. 496-7). For instance, "Twitter use still skews to younger, more urban demographic groups, even in wealthy nations like the US. In effect, this means older, less affluent and more vulnerable communities are often the least likely to be self-representing on a platform like Twitter" (Crawford and Finn 2015. p. 496). Thus, it is hard to obtain VGI data that is truly representative of the population being studied.
Finally, Crawford and Finn identify the limitations posed by the role of the VGI platform itself: it can be difficult to disentangle people's actual sentiments from the ways in which a platform like Twitter influences how they express themselves (p. 495). According to Crawford and Finn, "[Twitter] is not a neutral platform for observing people in a natural communicative space, nor is it divorced from the lived experience of a crisis. Instead, it is a space which has its own specific correspondence to the unfolding of a material event" (pp. 495-6). Further muddying the waters is the issue of bots: not all VGI from a platform like Twitter was actually created by humans, and it can be difficult to determine which content came from a person and which came from an algorithm (Crawford and Finn, p. 496).
Although uncertainty is unavoidable in almost all data and research, it is, as Crawford and Finn argue, the responsibility of the researcher to honestly and comprehensively acknowledge sources of uncertainty in their research. As Longley et al. put it concisely, "Uncertainties in data lead to uncertainties in the results of analysis" (p. 144). In order to accurately interpret results, one must understand the context in which they were generated, and part of this context is the potential for error. Furthermore, there is the very real ethical concern that those who seem to volunteer geographic information may not have truly given their consent for that information to be shared (Crawford and Finn 2015, p. 498). For instance, many people who texted the Mission 4636 SMS code to report information related to the 2010 earthquake in Haiti did not know that the data they shared would be publicly disseminated (Crawford and Finn 2015, pp. 494-5). As Crawford and Finn argue, using volunteered information for natural disaster research is not an excuse for violating the privacy rights of those who volunteer it: "The risks of big data approaches cannot always be known in advance, and communities experiencing a disaster should not be further exposed to potential harm without their consent" (p. 499). Therefore researchers should be very careful to use only VGI from people who consent to have their information disseminated. This complicates uncertainty because it potentially limits the scope of VGI that researchers can ethically use.
References:
Crawford, K., & Finn, M. (2015). The limits of crisis data: Analytical and ethical challenges of using social and mobile data to understand disasters. GeoJournal, 80(4), 491–502. https://doi.org/10.1007/s10708-014-9597-z
Longley, P. A., M. F. Goodchild, D. J. Maguire, and D. W. Rhind. 2008. Geographical information systems and science 2nd ed. Chichester: Wiley.