- April
Posted By : Adrian Chan
Hashing through twitter hashtags — a look at structured conversation

In my ongoing binge to ferret out the social mechanics of twitter third party sites and tools, hashtags deserves attention. It’s small, and by all appearances might die on the vine, which would be sad. But if the great culling whose season draws near were to remove hashtags from the social media dna pool, it would be for reasons owing less to the operationaI think and more to fundamental problems with the user experience.

Hashtags make a big demand on the twitterer. They ask that she tag up posts while writing, which requires a) added effort, b) a pause to reflect on meta while composing, c) a sense of the benefit provided to folks who search by tag in the future. (A) and (b) pull the user out of the immediacy of twittering — not much, but maybe enough to matter.

I remember working with a startup that was into the idea of tagging up chats for better discovery — of like-minded chatters and of topically-related chats. I thought the idea was great, especially because chats aren’t logged, and in theory at least the promise of social web is to connect people around what they talk about (in common). But to get meta from talk requires either automation or a change of user behavior. Either sites and services mine talk for the meta, and build links and suggestions of topics and talkers, or the user declares meta while talking (or just after, as in tagging).

Chat was one problem, twitter is similar. Because twitter is a kind of slow-moving chat. And like all other feed-related sites and tools, twitter I think aims to do more than provide dis-intermediated chat channels. Unless I’m completely mistaken, there are a lot of folks out there who believe that there’s value in fast web conversations, call it feeds, micro-blogging, or what have you.

And because it’s always hard to structure a conversation and to keep it on topic, not to mention expect users to follow topical threads, the goal of hashtags was to capitalize on what’s said *inside* of posts. Their benefit would be to extract topical continuity and consistency from disparate flows (conversations) — to re-flow if you will, by tag.

The benefit of exposing and surfacing common topical threads from talk accrues primarily to those who come along after the posts have been published. That’s the purpose of topical mining, and it’s an approach to making the web useful that Google has mastered for pages. But for posts, or conversational turns, the challenge is huge.

Conversation is notoriously poor at providing explicit meta about itself. It’s just not how it works or how we talk. We know what we’re talking about when we talk, and it involves the person(s) with whom we’re talking. We don’t declare what we’re talking about if it can be already understood. Seo is built on this — embedding meta in copy and pages for better search results. If you’re talking about orchids you’re unlikely to speak “flowers” in the same breath. We don’t supply meta while talking (Dave Sifry used to describe this as getting meta out of the exhaust).

A lot of social web sites live on the principle of using a small number of active participants to produce content that can be enjoyed by those who come later. The good ones, like Yelp, offload the member-to-member communication from the topical content as much as possible. In Yelp’s case, with gestural tokens, a questions page, etc — to keep the reviews as clean as possible. (Could tweeted reviews work even? They’d devolve into highly subjective recommendations probably.)

Two challenges face hashtags. First is getting the user to tag as an act of talking. Second is that all tags are equal. There’s neither vertical hierarchy (tag, subtag, subsubtag) nor modal organization. The latter is interesting, and worth a note, and since I just made it up I need to think out loud here a minute.

Talk is not all equal. Statements of speech can be declarative, performative, can be a form of request or appeal, and so on. Just think of the differences in the kind of utterance that are: invitation, recommendation, flirt, factual declaration, opinion, observation, question, answer. Modal organization of tweets would suggest that in addition to semantic tagging we have statement types: question, recommendation, announcement, link, flirt, and so on. You can see where this goes: Yahoo has Answers, Facebook has social apps, Linkedin has recommendations, Yelp has symbolic tokens, Google has (wait, google doesnt seem to get social. oops), etc etc.

So on hashtags, if you root around, you’ll find some totally unorganized examples of this. There are (and oddly or not so oddly enough Germans seem to have gotten into organizing talk — I lived in Berlin for a year and a half so I’ve had fun comparing German and English equivalents…). On hashtags there are posts tagged “now,” “love” “jokes” — tags that suggest the modal type of tweet. “Now” is not a topic but leads to conversations about what users are doing. “Love” is as much an expressive identifier (I’m in love, who’s in love?) as much as it is a factual identifier. “Jokes” is topical, but it also is an attractor for quips and humor — and joking is a performative form of speech unto itself if anything is.

When social media work to produce informative content for those who primarily consume, it’s because they are a means of production. By analogy to the industrial age, the communication age (our era) uses information technologies as means of production. Communication acts leave behind content. Interaction tools that simultaneously publish to a public, as are all social media, “manufacture” a new form of talk. It’s mass media gone social. But we know this already.

Is there a soft or hard threshold, then, for social media tools that ask the user to produce meta while in the act of posting/talking? Believers in the public utility of the social web as a means of democratizing the media and information might say no. Believers in connectivity, communication, and the social utility of social media might say yes. Hashtags sits between the two.

Much of my work involves helping social media companies to engineer interactions and communication for the purpose of producing leftover utility to latecomers and consumers. It’s a social engineering challenge — call it the art of generating unintended utility out of useful socially practices. It’s an art because it requires engaging the user’s motivation. Incentives (to benefit consumers by writing restaurant reviews) only work for a handful of users. For most, the byproduct of added value has to be in the exhaust — and the act of participation should be enough in itself to motivate engagement. Users simply have to like using the tools — for their own reasons (which are many indeed, but that’s a whole different story).

Let’s quickly take a look at this from a different angle — social media marketing. Twitter offers the promise of mining and tracking conversations, if not participating in them, with an unprecedented degree of proximity, directness, and immediacy. Twitter search engines and meme trackers currently offer more breadth than hashtags for the simple reason that many more users simply type “movies” than type “#movies.” But check “movies” on a site like summize (say you’re a social media marketer), and results will include this: “Uncle. I can’t work on this paper anymore. Let’s go to the movies!.” Use hashtags and you’ll get better posts, though far fewer, because users have declared their topical selection of “movies.” Intent is among the metrics measured by those in social analytics, and you can’t get better than hashtags for that.

If only users could be asked to declare their intentions while conversing so that mining companies could extract the gold! No, it seems far more likely that the success stories and serendipitous moments ahead for web 2.0 will be found among those tools that can engineer social interactions such that the meta is an accurate but unintended byproduct of talk and engagement. (Beacon?) We’ll have to watch this unfold and keep testing the frontiers. Either way, there’s a lot to talk about.


Leave a Reply