Hurricane Sandy is bearing down with full force on the east coast of the United States, and as I read Jack Dorsey’s tweet:

I knew that Jack has a lot to be proud of. I turned to Tweetdeck and took a look at the hashtag #Sandy and was blown away. Tweetdeck indicated that my stream was being throttled but still the stream for the hashtag was moving at a blistering pace. It was impossible to read a single tweet. This did not surprise me, among other things, Twitter is a personal broadcast system. A method to get a message out to the world and during emergencies, people want to get the word out.

In fact, Twitter’s roots come from Jack’s work with Emergency dispatch. I shut off real-time streaming so I could read some tweets and that is when I was able to get a feel for the situation. I noticed a few things:

1. People were in need of help and using Twitter to try to get help.
2. There was an amazing number tweets about the same things – the repetition inundated the stream.
3. Government agencies and emergency responders were trying to get safety information to those who needed it

One tweet by Newark Mayor Cory Booker spoke of a non-emergency phone number that was not working and he gave alternative numbers. This is more important than you might think. If you need assistance and you call the non-emergency and can’t get through you just might opt to call 911 next. This puts undue load on 911 and makes it harder for everyone to do their jobs. So the fact that Cory Tweeted this was great but what if you don’t know to check Cory’s tweets stream?

This is illustrative of one of the biggest challenges of using Twitter in emergencies like this. Making sure the important messages get heard and responded to. While individuals like Emily Rahimi, a seven-year veteran at the FDNY did great work at fielding tweets from people who could not get through to 911, the Sandy hashtag was moving too quick for any person or army of people to process. Not only was it moving fast but it was overrun by repetition and corrupted by disinformation.

This scenario is not unique to Sandy, it is typical for any large scale emergency and will only get worse as Twitter gains greater and greater adoption. The need to efficiently and effectively extract “critical signal” from the stream is clearly needed. Perhaps the answer is to apply Artificial Intelligence, informed by natural language processing and machine learning to the problem. This sounds like sci-fi stuff right? Well, no. If Watson can play Jeopardy, surely a system should be able to monitor a tweet stream for “critical signal”. Not only could this be used by FEMA and local gov’t to identify people in need, but it could also be used in some case to crowdsource help.
Take this tweet for example:

I saw it but I am in California. Could Artificial Intelligence help to identify someone from Lower Manhattan who tweeted that they bought a generator recently and get a message to them? Maybe. This is certainly a goal worth working towards.

Unfortunately emergencies like this are part of our lives and this will not be the last. I hope that before the next one we will realize that there is considerable “critical signal” that is shared on Twitter during these times and get a system in place capable of efficiently and effectively identifying that signal and acting on it.