Research

Posted 10 March, 2015 by Alix Dunn

Teaching the computer: how we are testing humanitarian machine learning tech with our advocacy partners

X-4 program with what Langley engineers euphemistically called "Female Computer" support personnel.

(Credit: NASA)

A big part of the engine room’s Matchbox project is to match the right experts with partners at the right time. Some of that requires networks and project design clarity, and sometimes it requires a little bit of serendipity. We were fortunate to come across an innovative project at the Qatar Research Computing Institute (QCRI) that was initially designed for the humanitarian sector. The tool, called Artificial Intelligence for Disaster Response (AIDR), aims to support humanitarian responders who want the help of computers to filter the firehose of tweets and SMSs in crisis situations.

So what does AIDR have to do with transparency and accountability projects? Well our partner Kubatana has done an amazing job of building a large community of engaged citizens in a WhatsApp group, and has been looking for ways to process incoming messages about water availability. They first want to be able to toss out the spam and then will be looking for ways to filter and organize the information in an increasingly efficient way.

We got in touch with Patrick Meier from both QCRI and his great blog iRevolution to see if AIDR might be right for Kubatana. After a bit of back and forth emailing, we were up and ready to go with an AIDR system to test. We’ll report back about how AIDR has worked for our partner, but we were excited to share this update (and thank Patrick). You can read more about AIDR here and here.

2 thoughts on “Teaching the computer: how we are testing humanitarian machine learning tech with our advocacy partners”

Patrick Meier says:

Thanks for experimenting with us by testing the use of the AIDR prototype for SMS classification. We look forward to learning with you.

Alix Dunn says:

Thanks, Patrick! We really appreciate how helpful you’ve been thus far and are excited to see if the AIDR can help solve some of these pesky firehose problems.

Leave a Reply

Your email address will not be published. Required fields are marked *

Related articles