I’ve been moved by the devastation wrought by the Haitian earthquake. It’s simply impossible to fathom, with death toll estimates hitting 200,000. In comparison, the Indonesian tsunami of 2004 killed nearly 230,000 people — placing it fourth among the world’s deadliest earthquakes. To give some perspective to those numbers, the atom bomb dropped on Hiroshima in 1945 killed 80,000 people instantly. These are numbers that I simply can’t grasp.
And this disaster still unfolds, with scores pitching in — many turning to the social web and social media to facilitate or amplify their efforts.
One such effort is being lead by Project EPIC, a collection of information scientists, computer scientists and computational linguists at the University of Colorado at Boulder and the University of California, Irvine.
I applaud their efforts and desire to help people communicate their status in a way that facilitates machine-processing. I worry, however, that this approach may limit its success.
Hashtags are metadata for humans first, machines second
The original need for hashtags came from the lack of any formal or public grouping mechanism in Twitter.
For example, when half of Silicon Valley went to SXSW and tweeted for days on end about this speaker or that panel, those who weren’t at the conference desperately wanted some way to filter out such noise. I proposed the hashmark (#) as a way of adding context to a tweet, so that people could choose for themselves to filter out or follow tweets tagged with certain keywords. In July last year, Twitter decided to hyperlink hashtags to their respective search results, and the format became widely adopted — more often than not used to game the trending topics on Twitter’s homepage.
Initially, most people thought hashtags were ugly and useless; even the folks at Twitter thought that they were unnecessary because they’d eventually develop natural language processing algorithms that would supersede the need manual tagging. Contrary to initial complaints about their complexity, hashtags become easier to understand and use with repeated exposure and practice because they are so transparent: if you see someone use a hashtag, you know how to use a hashtag.
And so three years later, hashtags still serve a role in helping people express themselves to each other.
Keep it simple, make it memorable
Language is inherently mutable; mathematics (the language of machines) is not. Verbal language can be adapted by a speaker, and what is heard (or read) is itself interpreted; the conversion is never digital, and invariably bears some loss of meaning.
But using hashtags to clarify meaning prioritizes the needs of the machine over the capabilities of the individual.
Such imposed order in a networked environment can succeed, but only if it achieves instant, widespread adoption, and is itself superficial (that is, it doesn’t require deep knowledge to understand or use the new order). In contrast, simpler, smaller and emergent structures tend to fare better over time, but developing them is not easy (see also: slashtags).
Successful structures should also aim for minimal cognitive burden — by being easy to remember and recall in practice. I’ve frequently seen people tweet about how they “forget to use hashtags” in posts — which is not surprising, since most people don’t think about the metadata of what they say. Hashtags and slashtags are most useful, therefore, when you want to provide additional context that is harder to express otherwise.
Learning from previous efforts
The Tweak the Tweet project introduces a “new order” for using Twitter. Though the words it calls out are mostly common, the use of the hashmark seems gratuitous, given the limited length of the medium (something that Stowe Boyd points out) and that the hashed words comprise the meat of the message, rather than the meta. To give you an example, this is Tweak-the-Tweet formatted post (77 characters):
#haiti #offering #volunteers #translators #loc Florida #contact @FranceGlobal
The same message could be reformatted to be human-readable without any loss of meaning (72 characters):
Offering volunteer translators in Florida. Contact @FranceGlobal. #haiti
While the message may not be as machine-friendly, it may reach a wider (human) audience available to respond to this offer.
Now, I don’t want to dismiss this effort, but instead provide a word of caution on focus. Tweak the Tweet is not the first hashtag pidgin language I’ve seen — and previous efforts struggled to gain adoption and awareness. Perhaps by minimizing the metadata and maximizing the meat, the effort poured into this might achieve a greater effect.
Paving the cowpaths and bulldozing fields
Hashtags may never have taken off if it weren’t for Nate Ritter tweeting about the San Diego forest fire in 2007. In fact, his use of the hashtag was the first dedicated use of a hashtag to help coordinate a response to a natural disaster:
What’s important about his use of hashtags in this case was that he was using them to communicate critical information to people in natural language. His use of the hashtag provided additional context to his followers who weren’t in San Diego, and also modeled a behavior that others could easily emulate when reporting their own news.
When I proposed using #sandiegofire as the hashtag for Nate to use, I first looked at what people were already using the tag their photos of the event on Flickr. At the time, the sandiegofire was one of the trending tags, and that’s how I chose it:
Had I tried to come up with my own new phrase for the event, Nate’s use of the tag may not have been picked up. #sandiegofire was also better than the alternatives, which were more localized and therefore more obscure to the broader audience. Using “SanDiego” in the tag itself helped bring clarity and context to Nate’s tweets.
Using hashtags effectively means considering the audience and their familiarity with the issue being tweeted about. While tagging lets you be as esoteric as you want, it may limit the reach of your effort, whereas paving the cowpaths means that you build on the familiar and connect with what people already know, reducing friction and inviting contribution.
iList with #ihave and #iwant
iList is an interesting service that originally aimed to take on eBay and Craigslist by leveraging social media. More recently they decided to narrow their efforts to focus on hashtag-based listings and Twitter search. Nonetheless, what I think is interesting about their approach is that it is, on the surface, quite simple.
To use the service, you just tag your tweet with #ihave or #iwant. If you want to get more detailed, you can add your zip code or categories like #forsale or #electronics. But the core service relies on using just two tags which seem to be have moderate usage — proving that getting adoption is always the hard part of any metadata-based communication strategy.
Twitter Vote Report#votereport
The last example is very similar to Tweak the Tweet and was launched by some friends of mine. The Twitter Vote Report project was designed to enable citizens to report on their local voting situation by using a series of hashtags:
- #[zip code] to indicate the zip code where you’re voting; ex., “#12345?
- L:[address or city] to drill down to your exact location; ex. “L:1600 Pennsylvania Avenue DC”
- #machine for machine problems; ex., “#machine broken, using prov. ballot”
- #reg for registration troubles; ex., “#reg I wasn’t on the rolls”
- #wait:[minutes] for long lines; ex., “#wait:120 and I’m coming back later”
- #early if you’re voting before November 4th
- #good or #bad to give a quick sense of your overall experience
- #EP[your state] if you have a serious problem and need help from the Election Protection coalition; ex., #EPOH
All tags were optional except the #votereport tag.
They also went through painstaking effort to mobilize people and provide alternative means to participate. They also did a good deal of work to report back their findings in real time (most visualizations appear to be offline) and open sourced their codebase.
They also made sure to make it possible to participate without using Twitter — the hashtags were just a mechanism for getting data into the system.
Design for adoption, stay focused
Around the time it launched, Ethan Zuckerman expressed skepticism about whether Twitter was the appropriate tool for the vote report project, in much the same way I’m wondering whether Tweak the Tweet could take a more focused approach in exchange for wider participation to achieve its goals.
My greatest concern is that there won’t be enough people who can “speak” the “tweaked” syntax, leading to a lot of effort spent building parsers that will be data-starved. While trained volunteers might be able to use this syntax effectively, I wonder if there aren’t alternative approaches that could use the existing corpus of text messages and tweets coming out of Haiti (which probably aren’t geo-coded, unfortunately) to discern the typing patterns that people use naturally in order to facilitate adoption? Perhaps by focusing on fewer tags that are self-evident in their meaning and use, it is possible that this effort could be used to model the proper usage of the tags, making a more direct difference while there’s still time? Unless the audience of this effort is expert users, I’d suggest steering towards simplicity and ease of adoption — and being mindful that typing out a complicated machine-friendly syntax might be the last thing on someone’s mind who’s trying to find or offer help in such a disaster.