Of course Twitter is conversation

A couple of weeks ago, Mark Drapeau wrote a post that alleged that Twitter was not a tool for conversation, but for broadcast. It’s a provocative point, and is clearly false. Twitter isn’t a very good medium for extended conversation – but it’s obviously used for both conversation and broadcast.

The article uses statistics about the number of posts per Twitter account to infer that most Twitter activity is publishing. This isn’t a good interpretation of the facts for a couple of reasons. The low number of posts per account is almost surely evidence of a high rate of “dabbler” use. People sign up for Twitter, look around, and go away. The data about number of posts per account doesn’t say anything about people who are active on Twitter but use it primarily to consume content produced by others. There isn’t any evidence about the relative ratio of reading vs writing.

The second misreading relies on the Pareto principle – the highest volume of posts comes from a few people. This is true but irrelevant. Let’s say CNN has a service that publishes 100 updates per day on news stories. And two people have a conversation consisting of 5 posts each. These are two different, valid use cases. The existence of high-volume broadcast messages doesn’t somehow negate the fact that some people are talking to each other.

Direct evidence that that Twitter is conversation can be seen in Tweet Tweet Retweet a research paper by danah boyd and fellow researchers studying the use of Twitter. According to the paper, “36% of tweets mention a user in the form ‘@user’; 86% of tweets with @user begin with @user and are presumably a directed @reply.” The data uses on “a random sample of 720,000 tweets captured at 5-minute intervals from the public timeline over the period 1/26/09-6/13/09 using the Twitter API. This sample includes tweets from 437,708 unique users.” Another study with over 1 million tweets shows the same pattern – 39% tweets have an @user mention and 19% contain questions. (Thanks, Juan Carlos Muriente, founder of )

That looks like conclusive proof of the conversational use of Twitter. This surely dovetails with my own experience. In the last week, I’ve had conversations on distributed social networks, music, and Bay Area public transit. In these conversations I learned new information, met new people, shared ideas, and set the stage for follow-on activity. Twitter works for conversation, and the open nature of twitter sparks conversations that might not occur otherwise. It is true that Twitter is a not a good medium for in-depth, extended conversations. Messages are restricted to 140 characters. There isn’t visible threading (although thread info is kept in the data, allowing for threaded views such as Tweetboard.) The richest conversations sparked by Twitter often take place on Friendfeed, where replies are threaded in FriendFeed.

Twitter is good for short, fun and/or productive conversations that bring in often-unexpected relevant people through the social network. Deeper conversation and deeper collaboration need to segue into other modes. The next frontier for development, being pushed in different ways Google Wave,Citability, and other tools and concepts, will be means to connect shorter, real-time conversations with more in depth conversation and collaboration.

5 thoughts on “Of course Twitter is conversation”

  1. Hi Adina,
    I’m guessing that Mark Drapeau doesn’t really talk with people that much. His post seems to pick stats that back up his assertion that twitter is more like wikipedia, instead conversational. I think microblogging is a lot like blogging but smaller:

    * 10% of the people may do it the most but if they are talking to each other often, even asynchronously, who cares? it’s still conversational

    * twitter doesn’t keep your info longterm.. which makes it not at all like wikipedia

    * searches can aggregate a topic, but going in reverse chron view makes it’s also not like a wiki, but like a conversation

    I think danah and others are right on this.. twitter is conversational, people drift in and out of conversations, and they spend time directing their words to specific people because of the @’s in their tweets.


  2. I operate a bunch of Twitter accounts — business, personal and community-related. For what it’s worth, some work mostly in broadcast mode, some lean strongly conversational and for some I actually make use of both. And from what I’m seeing, that’s not entirely unusual either. To say that Twitter is not conversational seems a bit of a stretch then.

  3. You seem to define “conversation” in a very casual manner, as in, when people send information to other people, that’s a conversation. If you read my post, I define it along the lines of what Clay Shirky wrote in his book, that a conversation meaningfully occurs when a defined group of two or more cooperates with each other in ways that form some group identity. True, meaningful conversations don’t happen when no one knows who’s in some undefined, asynchronous mess of information. Info is shared, perhaps knowledge is created, but very little real conversation happens.

    When I write to @andersoncooper and say “Great show today!” and he doesn’t notice, that’s not a conversation. When a bot account writes @cheeky_geeky and auto-replies “Thank you for tweeting about vaccuums! Would you like more information?” that’s not a conversation either. When someone says RT @scobleizer OMG he is giving a killer keynote! that’s not a conversation. And when @juliaallison twitpics herself with the note “Am I hot in this dress? @ me!!!” and while everyone is writing her, she’s on a date, that’s not a conversation.

    Saying that 35-40% of tweets have an @username in them means “conclusive proof of the conversational use of Twitter” is ridiculous.

  4. It is true that only sending is not evidence of conversation. In particular, @addressing by bots is surely not conversation, and a more rigorous methodology would filter out bots.

    As for @address and @mention of people, there are some other techniques that could refine my hypothesis that most of this actually is people talking to each other. If the other person replies back, that’s probably conversation. But even if the person does not reply immediately, but the people regularly exchange communications, that too is likely to be conversation, of an asynschronous and casual fashion. (and is where my definition varies from yours)

    It seems to me that assuming that people who address each other 36%-39% of the time are actually talking to each other is the most plausible explanation based on common sense experience with people, as well as experience of Twitter.

    Assuming that people addressing each other are *not* talking to each other *could* be true, but imagines an autistic world that does not seem like the most plausible explanation.

  5. The definition that a conversation needs a group identity and shared goal sounds excessively rigorous, depending on how one defines group identity and shared goal. For example, I tweeted last weekend about buddypress. Three different people replied to me, and I then engaged them back with questions & comments. There wasn’t a pre-existing “group.” One notable moment happened when a friend introduced a colleague with a common interest to me, based on the posts. And I didn’t have a particular goal then other than sharing my experience. But I know that on Twitter sharing experience in my network often leads to rewarding exchanges, too, and it happened again here.

    In particular, in Twitter, the lack of a pre-defined group and pre-defined goal is particularly helpful and delightful. But it doesn’t mean it’s not conversation. I’ll have to go back and read the Shirky definition in context to see what he was trying to get at. I doubt he’d think that Twitter exchanges aren’t conversation. I’ll write another blog post and ask him :-)

Leave a Reply

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

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <strike> <strong>