Humanitarian initiative MicroMappers relies on a global, online network of volunteers to help sort through the piles of Big Data – including tweets, texts, images and videos – that amass online during times of crisis. The vast amount of information posted to a site such as Twitter during a disaster can be overwhelming, but with analysis and classification of the data, it can help create a more holistic picture of what’s happening on the ground .
Take for instance MicroMappers’ response to Typhoon Yolanda (Haiyan) in the Philippines in 2013. In that case, the UN Office for the Coordination of Humanitarian Affairs (OCHA) had specifically requested support. Preliminary figures showed that in the first 48 hours after the typhoon, volunteers sorted through 30,000 tweets, tagging them based on relevancy and urgency.
Similarly, volunteers looked at photos that had been tweeted and rated the level of damage they saw in each. Those tagged “mild” and “severe” were then geolocated on a live map, so that first responders could identify urgent needs.
These volunteers, dubbed “Digital Jedis,” work through computer-based applications known as “clickers” to help people sort through the piles of data.
Digitial Jedis, who are part of the Standby Task Force, are alerted by MicroMappers when help is needed. The organization also draws on assistance from the standing Digital Humanitarian Network, co-founded by MicroMappers’ creator Patrick Meier and the UN in 2012 to connect digital volunteer networks.
“Today, with the DHN, we simply post a request for help on the DHN list-serve and also on social media when humanitarian organizations like the UN need our help to make sense of the Big Data generated during disasters,” Meier explained in an email.
For instance, someone might tweet a photo of infrastructure damage, which then appears in the Photo Clicker, where a Digital Jedi can classify the severity according to preset categories. The same general principle applies to tweets or SMS messages, whether they are from individuals in crisis or news agencies, about food, shelter, injured or deceased persons, or another aspect of the situation.
Each image or tweet is sent to three different Jedis.
MicroMappers also has a Video Clicker, Aerial Clicker and Geo Clicker, used to geotag already filtered information.
The group is also working on an Artificial Intelligence for Disaster Relief platform (AIDR). The goal is to speed up the intake of tweets by teaching a computer to accurately mimic the way volunteers classify information.
Using AIDR’s Collector tool, the tweets a volunteer sees can be filtered by language, region and according to certain tags, such as building damage or water infrastructure. As the Digital Jedi classifies text, the computer will learn how to do the same and start to classify tweets as they come in in real time at the rate of 30,000 tweets per minute.
AIDR assigns its classifications a confidence tag based on how certain the artificial intelligence tool is that it categorized the text accurately. Tags below a certain confidence threshold get sent to the Digital Jedi so the tool can learn how a human would interpret the text.
MicroMappers has broader intentions than just disaster response, however. It also worked with on a map tracking wildlife in Namibia for an anti-poaching initiative, one of six projects it has tackled.
The initiative isn’t without its critiques. It can’t reach everyone, being limited to cellphone (often smartphone) and social media users. Internet outages common to disasters can restrict the amount of available information.
Meier has countered that this is one response among many during crises and that they are aware of the limitations of it. They are able to use it to reach certain people, while another initiative would be needed to aid those who are excluded.
“Social media is simply another node in the pre-existing ecosystem of crisis information,” Meier explained in a blog post.