Letters: The problem with mapping hate speech

Last week, I wrote about Hatebase, a new project from the Canadian-based Sentinel Project, which invites users to submit both hate speech terms and examples of their use in order to track upticks in hateful rhetoric before outbreaks of ethnic violence.Radio Milles Collines infamous broadcasts before the 1994 Rwanda genocide are an example of a type of phenomenon the project's creators are looking to study. 

Gwyneth Sutherlin, a doctoral candidate at the University of Bradford whose writing focuses on "conflict resolution policy and tech solutions examined through a cultural lens" and is currently conducting field work in Uganda, sent in the following response:

In Acholi, the language spoken in Northern Uganda where Joseph Kony's LRA was recently driven out, as well as the new nation of South Sudan, there is no word for map. There are over 40 languages spoken in the region, and none have a word for map. Why then do most conflict prevention information and communications technology (ICT) applications organize information around maps?

The disconnect is deeper than language. It extends to the way in which cultures conceptualize information -- what is deemed worth capturing and how it should be organized. For example, most applications organize information around a linear concept of time. Some cultures have a circular concept of time, or even a combination of the linear and circular. If you find this hard to imagine, then you have begun to understand how someone from those perspectives must adapt to share information in the current format. Beyond changing the language users see on the interface, ICTs have not adapted to other cultural concepts of knowledge management. This has profound implications for whether or not local users have access to the information or if they feel it's relevant for their own policy implementation.

A recent War of Ideas post profiled the conflict early warning initiative called Hatebase which collects hate speech via a Wikipedia-like interface and correlates the entries with reports of violence. The assumption being that the Rwandan genocide was preventable and monitoring local language patterns could play a part. However, the feeling of being caught by surprise and failing to prevent the conflict are the feelings of an outsider, not those of a Rwandan. So it must be asked, who is this tool for? Individuals in conflict-affected areas are perpetually simplified into victims waiting to be given a voice. Rwandans had the information, it just wasn't in written form accessible to the West. The oral nature of information in other cultures has so far been discounted in ICT design; only when information takes written form is it cemented as fact, as actionable, as ready for analysis.

Among the nearly 6800 languages, only around 100 developed a literature, the rest remain oral to some extent. Orality in a culture does not imply underdevelopment or opposition to literacy, simply a preference for that mode of expression. There is enormous potential to re-imagine the visual interface, to respond to cognitive cues and communication norms from the cultures crowding into the digital space... Nigeria, Indonesia, Brazil, China. This is an unexplored avenue for ICT design.

How would Hatebase contend with the Acholi word gwok? It can be pronounced six different ways in order to mean either: dog, shoulder, incapable, misfortune, do not, or protect. You have to hear it. Information is something you hear, collect and pass on. Gam means both to receive answers from someone (and pass them on) and collect people in a taxi and drive them off. So it is with information here. It moves. It defies conventional categories. Hatebase may have research value, but without integrating more culture-specific concepts of how to collect and organize communication data, it will never gain a predictive capacity or relevance within conflict areas.

Quick note to readers: I'm happy to feature responses to any of the posts on the blog, particularly from topic experts. E-mail me at joshua.keating [at] foreignpolicy.com. Please keep them short, constructive, and respectful.