How not to use twitter in an emergency

Southern Alberta, including the City and County of Lethbridge, had an interesting afternoon. Fortunately, few if any people seem to have been hurt or lost their homes, though the fire may have hit at a terrible time for area farmers, who were in the middle of harvesting a bumper crop in a high-price market.

Lethbridge is not a major media market and is not represented on our satellite TV (though we could get lots of information about a storm affecting the East coast). Local (commercial) radio continued to play music through the emergency, much as the Soviets used to do when their leaders died, except with more Fleetwood Mac and NickelBack and less Tchaikovsky.

The most accurate and up-to-the-minute information was available through Twitter. Local newspaper and television reporters joined with the City, County, University, and province to tweet the latest situation updates and evacuation orders in what was, by definition, a rapidly developing situation (we were, after all, discussing something spreading like wildfire… non-metaphorically).

Numerous tweets reminded people to pay attention to the official channels of information. And one of the most important official channels of information was the province’s Emergency Alert feed. This was responsible for posting authoritative evacuation alerts and updates that were commonly passed on by the news agencies and local residents.

A very important set of updates during these hours involved the small town of Milk River. This is a small agricultural village of about 811 residents, located about 70km South East of Lethbridge (and 15 km north of the U.S. Border) that was threatened by a second, equally dangerous grass fire.

In contrast to the situation in Lethbridge County, the situation outside Milk River was extremely volatile–a fact reflected in the alerts published by AB_EmergAlerts. The first evacuation order for the town was reported at 6:40 PM: AB_EmergAlerts advised residents of Milk River to head about 17km SE for the border town of Coutts Alberta/Sweetgrass Montana. Twelve minutes later, they posted a number residents could call if they were unable to get themselves to the emergency shelter in Coutts.

A little over an hour later, however, the situation changed significantly and an urgent warning went out to residents of the village: due to a drastic change in wind direction, it was no longer safe to flee to the south; residents were now to head north to a centre that was being set up in the town of Raymond, 60km or so to the NW. Again a number was provided for residents who did not have transportation.

If you’ve been following this story by clicking on the links I’ve provided above, you will, at this point, have come to two conclusions:

  1. The precision of my timeline is not really that impressive, since each of the tweets I’ve linked to has two timestamps–one provided by Twitter and the other included in the body of the tweet from AB_EmergAlerts
  2. You have no direct way of checking the accuracy of my description of the actual  emergency instructions, since none of the tweets tell the residents of Milk River what they are supposed to do.

In fact, every tweet published by AB_EmergAlerts follows a similar template:

<broad characterisation of the nature of the emergency> “Updated” <time stamp> “Take necessary precautions.” <location>  #abemerg

Or, in this case:

Wildfire Alert Updated Sep10 0746PM Take necessary precautions. Milk River  #abemerg

A page showing the twitter feed for AB_EmergencyAlert on the evening of September 10, 2012

I am as big a fan of metadata as the next person, but I have to confess, placing myself in the shoes of a resident of Milk River heading towards Coutts to escape a prairie grass fire that has turned around and is now heading towards me, that I might be willing to sacrifice a little context for actual content in this case.

Twitter is a great way of providing up-to-the-moment information, but it only allows messages of 140 characters. Devoting 22, including spaces and punctuation, to a timestamp (something Twitter already provides), 5 to reminding me it is an alert (on a steam that I am presumably subscribing to because it advertises itself as being an “EMERGENCY BROADCAST FEED, CRITICAL ALERTS ONLY” and the “Official channel of Alberta emergency public alert system”), and 28 to the good if generic advice that I should “Take necessary precautions,” seems a bit much even for somebody who can barely suppress his inner librarian. When I am acting on evacuation instructions that have been superseded because the wind has changed direction and the wild fire I’m fleeing from is now heading towards me as quickly as I’m driving towards it, I think I’d be willing to sacrifice the redundant certainty that Alberta Emergency Alert issued revised instructions at 0746PM for the actual content of those same instructions (and I’d probably thank God that lived in the 10 character town of Milk River: I hate to think what information the staff at Head-Smashed-in Buffalo Jump would have had to sacrifice in order to have their 28 character location spelled out in the Tweet warning them to flee for their lives).

So what should you do if you are tweeting emergency messages to a population? Here are some basic rules:

1) Focus on the content, not the metadata

Twitter is now a primary source of emergency information and, in an emergency, the target audience is almost certainly mobile (i.e. accessing your feed on a mobile device). If there is critical information you must get to this audience, it should be the main focus of your 140 characters. If you can, get the important stuff in the headline.

If the information is too complex to present entirely in such a short space, or if there is other information that must also be presented (e.g. location, effective time, etc.), concentrate on making your tweet as actionable as possible on its own terms. “Milk River Evacuate to SE 746PM” is a far more effective use of your space than generic advice to “Take necessary precautions,” with detailed but redundant information about who you are, when you posted the message, and a link to you home page.

In other words, do not waste space on noise: Twitter already indicates who a message is from and when it was sent: if you are pressed for space, you can leave that information out, with little if any harm.

2) Keep it short

In addition to wasting space on unnecessary and redundant information, the AB-EmergAlert commits a number of cardinal errors in the information it does include. Consider, for example, the “Updated” time. In addition to wasting 7 characters on telling us that the information was “Updated” (something that was, in fact, not true on the first notification), AB-EmergAlert also wasted characters telling us that the update was posted at 0746PM.

As mentioned above, this information is actually not necessary, since Twitter time-stamps its messages. But even it if had been necessary, this is a maximally verbose way of supplying a time stamp. The most efficient, presumably, would have been to use a 24 hour clock: 1946 saves two characters or 1.5% of your life-and-death message. Even if we use AM and PM because we are worried that a general audience might be confused by the relatively rarely used (in North America) 24 hour clock, there is no excuse for the leading 0: there is no context is which 746PM might be preceded by anything other than a leading 0 and the time is perfectly understandable without it.

3) Never put any crucial information more than one click away from the tweet

The AB_EmergencyAlert splash page: Quick! Which way are we supposed to go?

In addition to filling their tweets with useless information, AB_EmergAlert compounded their audience’s problems by pointing at a website that was as indirect as the tweets that sent them there. Instead of sending citizens directly to the specific information they need to act on as they are fleeing the grass fire the wrong way on a smoke-obscured highway, AB_EmergAlert instead sent them to a page summarising the metadata for all the most recent tweets sent out by the agency. Meaning that while you may be still driving hell-for-leather towards the fire that AB-EmergAlerts would really like to warn you has now changed direction, you can rest assured that the good people of Lethbridge 70 km to the north were also given unspecified critical alerts about a completely different fire at 0512PM, 0420PM, and 0335PM. And oh yes:  that there was also something wrong, but not really critically important, about the power supply in Hanna, Alberta–an equally small town four hours (301km) to the North.

So how do we actually discover what was so important that AB_EmergAlert felt we should have an update at 0746PM? All you have to do is follow the following steps:

  1. Scan through the page of similar looking red boxes (if you took a moment to look at the legend at the top of the page, you will know that this is colour used to indicate “Immanent life-threatening danger”). Conveniently, these have been sorted by time rather than location.
  2. Looking at the second line of each box, find the one that refers to your town and, on the first line, the time that corresponds to the “updated” time on the tweet you just left in order to navigate to this page (hopefully you wrote that updated time down with the hand that wasn’t steering you towards the fire they want to warn you has changed direction before you clicked through).
  3. Look at the third line, for the hyperlink that says “More details.”
  4. Click on that.
  5. Look under the reporting history and to the left of the map (if you want, you can click on hide map to reload the page and make things easier to oversee).
  6. Read that the wind has changed and you are now supposed to be heading the other way as fast as you can.

4) Remember that people are reading your emergency notices on mobile devices under unfavorable conditions (i.e. during an emergency)

Trying to figure out what to do on a smart phone.I worked out the above sequence sitting in a chair and working with a desktop computer far away from Milk River and long after the emergency had passed. Presumably, if I were fleeing the grass fire in the wrong direction, I would not have any of these luxuries: I would be looking at the tweet on a mobile phone with poor internet service (the Milk River-Coutts corridor, like a lot of rural Alberta, does not necessarily have big-city mobile coverage). Fortunately, AB_EmergAlert decided not to deprive such users of the full desktop experience just because they are fleeing for their lives and using a small screen. As the following screen shot, taken on a brand new Samsung S3 Galaxy shows, people who were on the road but, having made it this far, nevertheless wanted to discover precise details of the “Immanent life-threatening danger” AB_EmergAlert was tweeting about had the difficulty of their work compounded by the fact that page they were directed to was optimised for a desktop computer: not only do you have to find the box referring to your particular life-threatening emergency and click on the generic link that will ultimately take you to the information AB_EmergAlert thinks you really need to know right now, you need to do this with text that is about 1.5pt tall. (To be fair, they have thought of this: you can also scroll to the top of the page, click on the “Mobile version” link, wait for the page to reload on your spotty mobile service, and then repeat the process of trying to find the precise urgent message AB_EmergAlert wanted you to know at their 0746PM update).

The grass fire was an interesting experience because it showed the power of the new social media to broadcast important (and potentially life-saving) information widely and quickly during an evolving emergency. And fortunately, while it looks like significant economic damage may have been done in the region, it appears at this moment as if nobody has lost their life or been seriously hurt.

But while social media provides a useful way of broadcasting emergency information, its success is only as good as the use you make of it. In establishing an emergency twitter feed, it is crucially important to maximise the efficiency of the information you present and to consider the real-life circumstances under which it is being used. Good emergency management is a rhetorical and generic problem as much as one of logistics and equipment.


Get every new post delivered to your Inbox

Join other followers: