Collection of Open Source GIScience Projects
When formulating a response to something as consequential as a natural disaster, it is necessary to consider the limitations of crowd-sourced/volunteered geographic information (VGI) data. Power outages, lost connections, or overwhelming situations can lead to only a partial representation of the scope of a disaster scenario. The use of VGI data raises important questions of representation, and, as we saw in the Wang et al. (2016) reading, there exist hierarchies of who consistently gets shared/retweeted/etc. (i.e. who has a voice) on social media sites. There are also power relations regarding who generates disaster data and who ends up being able to use it. This hierarchy of data access was prominent the Mission 4636 response network in Haiti discussed in Crawford and Finn (2014) and the ethical concerns surrounding this dynamic are complicated by the limited awareness of those involved in generating the data regarding how it is to be used. In the Haiti example, a language barrier also comes into play post-translation to English, and this combined with data permissions or costs raises important questions of data access.
As we discussed in class, user-generated information can be empowering. However, there is also high potential for untruthfulness in this type of data (intentional or not), and also high potential for ambiguity. VGI data may be used to create imperfect indicators to study a phenomena, which can introduce uncertainty into an analysis (Longely et al., 2008). VGI data also may introduce uncertainty in the form of positional accuracy when tweet content rather than location services/coordinates are used to geolocate where a post is coming from. Additionally, not all demographics are well-represented in these data; rather, there is typically a high representation of young, wealthy, urban-dwelling individuals (Crawford and Finn, 2014). The issue of bot accounts must also be considered. When using VGI data, it must be acknowledged that it is incomplete rather than assuming that it provides a perfect representation of a situation.
Additionally, Crawford and Finn (2014) bring up an excellent point: apps like Twitter are not neutral spaces. The practice of tweeting is influenced by the development of an event and the content of the tweets others are posting or retweeting. This introduces bias and feeds into differences in representation. Further, there is generally no explicit separation of emotion and fact in the contents of a tweet. These forms of uncertainty can be classified as uncertainties of conception (Longley et al., 2008).
There is certainly an interesting question of privacy and what is revealed when using VGI. The sharing of location data and information revealing “vulnerability” during crisis events could expose personal sensitive information. The authors’ discussion of the context in which information is shared raises important questions about situational priorities and how the effects of actions in difficult moments may linger long after a disaster “event” has occurred. The “for the greater good” approach to crisis mapping seems to muffle these concerns, and also those concerning representation or biases in the data (Crawford and Finn, 2014). Ultimately, this can result in “forced consent” with regard to data sharing.
Readings:
Crawford, K., and M. Finn. 2014. The limits of crisis data: analytical and ethical challenges of using social and mobile data to understand disasters. GeoJournal 80 (4):491–502. DOI: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. Print.
Wang, Z., X. Ye, and M. H. Tsou. 2016. Spatial, temporal, and content analysis of Twitter for wildfire hazards. Natural Hazards 83 (1):523–540. https://github.com/GIS4DEV/literature/blob/master/Spatial%20%2C%20temporal%20%2C%20and%20content%20analysis%20of%20Twitter.pdf