Posted by: structureofnews | March 23, 2012

Getting Personal

There’s a great piece in Slate that takes the notion of personalization on the web to its logical extreme: The individual reading experience for every story, for every person.

And why not?  After all, we can already – in theory and to some extent in practice – customize machine-generated stories and reports by drawing on different information in databases (“Your stock portfolio tanked today despite the overall market rising….”).  It’s not that far a step to imagine we can also use knowledge about you and your online habits to craft a tale that appeals to your interests and knowledge.

Is this a good thing?  Perhaps.  Although not everyone will agree.

Evgeny Morozov, who wrote the Slate piece, worries about the social impact of a world where we all read slightly – or very – different stories about the same event.  And about the increasing power of companies like Google and Amazon that will have far greater capability to do this successfully than other media organizations, by dint of the horde of data they’re gathering on our habits.

If there is one unambiguous trend in how the Internet is developing today, it’s the drive toward the personalization of our online experience. Everything we click, read, search, and watch online is increasingly the result of some delicate optimization effort, whereby our previous clicks, searches, “likes,” purchases, and interactions determine what appears in our browsers and apps.

Imagine that my online history suggests that I hold an advanced degree and that I spend a lot of time on the websites of the Economist or the New York Review of Books; as a result, I get to see a more sophisticated, challenging, and informative version of the same story than my USA Today-reading neighbor. If one can infer that I’m also interested in international news and global justice, a computer-generated news article about Angelina Jolie might end by mentioning her new film about the war in Bosnia. My celebrity-obsessed neighbor, on the other hand, would see the same story end with some useless gossipy tidbit about Brad Pitt.

All of these are real concerns – and that future may not be as far off.  The quality of machine-generated stories is steadily improving, and the world is increasingly awash in the kind of data that such programs need to create – and personalize – pieces.

Should we worry about a world where we read stories tailored to our interests – and prejudices?  Probably – although to some extent that already happens, through the choice of what news organization we go to.

But there are also compelling reasons to welcome this world as well.

For one thing, it can help make news really matter to readers.  If I’m going to read a story about how the stock market performed today, I’d certainly prefer it to note, reasonably high up in the piece, whether any stocks in my portfolio moved significantly.  When I read about hospital rankings, I’d like to see how the one in my neighborhood fared.  If there’s a piece about city hall corruption, I want to know if my elected representative is one of the bad guys.

And I don’t really want to have to dig through a database just to find out just those specific facts.

Right now, we write a story about the broad trends, and then push people to look up the details that they really care about elsewhere – in the first example, checking stock listings.  Why should we do that?  That’s condemning a trove of valuable information to what Matt Waite once called the “data ghetto.”  Let’s instead surface it in a way that lets people access and understand it easily.  If they want to dig through the database as well, let’s let them do that too.  But we shouldn’t force them to.

The main reason we haven’t done it before, of course, is that we couldn’t.  It was physically impossible to write stories for every subscriber describing the situation in their portfolio.  But automation now allows us to.  Or at least it allows us to move in that direction.

Similarly, there have been experiments in tailoring the reading experience to a reader’s level of knowledge.  Google’s Living Stories experiment tracked your browsing and grayed out information you had read before.   That’s a small step, it’s true, but it’s something to potentially build on.  Who among us hasn’t been annoyed by reading for the nth time boilerplate background on a running story?

How might story selection change if such capabilities existed?  Would we skew towards chasing issues that we knew we had highly-customizable data for, knowing that we’d be creating highly personal stories for our readers? Would that be a bad thing or a good thing?

Regardless, for this to work, we’ll have to rethink some of the ways we write stories, breaking them into more atomized parts that can be allow for easy substitution, depending on who’s reading the piece and what datasets are being drawn from – not unlike some of the ideas embedded in structured journalism.  At the very least, we’d have to write some parts of such stories in a quasi-structured way, so that different data points can be easily swapped in and out.

(Eg: Overall, hospitals averaged a 80 out of a possible 100 points.  The hospital in your neighborhood scored 65, placing second to last among all 53 area hospitals.   Mortality rates at your hospital….)

We’d give up some flexibility in prose; but in return we’d be serving more readers better.  Not a bad trade.  And with better engagement – and a real reason for readers to subscribe and provide us information – we may well be able to better monetize the fruits of our labor.

Still, the world described in Slate – of perfect personalization – is probably a ways off.  But if we consider how successful Google has been in intuiting our needs from our behavior – and catering to us – it’s certainly not a complete fantasy.


  1. […] everybody will be able to have completely personalized news stories thanks to algorithms, huge databases of cliches and text robots. […]

  2. […] text will find its way into newsrooms in the near future: Lower costs, broader coverage, greater personalization.  But it may be that automated trawling for insights in large datasets is the most useful one in […]

  3. […] serialize stories.  We can build new functionalities that allow users to customize stories with personally-relevant data.  And so […]

  4. […]    話題のベンチャー企業「ナラティブ・サイエンス」が提供している記事の自動生成ソフトの紹介だ。30社ほどの顧客がいるといい、有名なのは雑誌「フォーブス」だ。企業の決算関連の記事作成に利用しているようだ。ソフトにデータを入力すると、ものの数秒でデータの読み解きも含めた記事が出来上がるのだという。  この自動生成記事を、さらに金融取引システムが読み解いて、自動取引に生かしていく様子は、「まさに、ロボットによるロボットのためのジャーナリズムだ」とモロゾフ氏。そして「唯一いい面があるとすれば、その結果、金を手にするのは人間だということか」  さらに、恐ろしいことが書いてある。自動生成ソフトは、安い(500語あたり10ドル)、早い(ものの数秒)、文句を言わない(笑)、そして数百万というツイートを解析し、記事化することも出来てしまう。  そしてこの傾向の先には、利用者のネット使用履歴を反映した極端な記事の「パーソナル化」が訪れるのではないか、と指摘している。同じ記事なのに人によって盛り込まれるデータが違う。そして、人々はその違いに気づかぬまま、思考はどんどんタコツボ化していく…と。  これに対して、記事の自動生成やパーソナル化は、そう悪いことではないのは、とブログで主張したのがトムソン・ロイターのデータ・イノベーション部長、レグ・チュア氏だ。  株価の記事を読むなら、自分の持ち株に急な動きがないかは知りたいし、病院のランキングの記事なら、近所の病院がどの当たりなのか知りたい。それに、市役所の汚職の記事なら地元選出の議員が絡んでいるのかどうかも――そう、チュア氏は指摘する。  これまでは、記事では一般的な傾向について書き、読者が具体的なデータを欲しい時には、(例えば株価なら商況欄を)自分で探して…という組み合わせだった。だが、読者がいちいちデータベースなどを探し回らなくても、個別に必要なデータを、分かりやすい形で盛り込んでいけばいいのでは。  これまでは技術的に不可能だったが、記事の自動生成技術を使えば、それが可能になってくる、とチュア氏。だが、そのためには、自動生成しやすいよう、データの入れ替えが可能なように、記事の書き方、組み立て方も変えていかなかれば、と言う。 […]

  5. […] before, and touched on the advantages it can confer, not least scale, speed, volume, cost and personalization.  But I confess I’ve never thought deeply about what it meant in news judgment […]

  6. […] speak to audiences of as small as one, opens up a raft of other potential uses, including much more personalization and the creation of “news-on-demand.”  And that opens up questions about what’s […]

  7. […] of structured information. But technology has advanced significantly so that the prospect of personalized narratives and exploratory “counterfactuals” and news-on-demand are fast becoming a […]

  8. […] while it’s not quite the kind of extreme personalization that news online is theoretically capable of, it’s yet another important step in that direction. And not so much just because these […]

  9. […] and when you get machines to write the stories, which I’ve argued is both the only way to scale personalized/on-demand story creation as well as take full advantage of a database of atomized facts. Even stunning, rich data […]

  10. […] once you have the parts of information that create news, from smart data visualizations to intense personalization of news, created on […]

  11. […] how my investments are faring (not to mention, is there an earthquake near me). And that’s where personalized news-on-demand can help fill a core public information need. It doesn’t have to be automated, of […]

  12. […] – the late-lamented Circa, or Google’s Living Stories – and something I’ve written a fair amount about, both as where journalism needs to be going and what structured journalism can help power.  And it […]

  13. […] other humans.  Duh.  But a human being can’t give hundreds of thousands of individuals a personalized newscast, or provide one on demand with the latest stories.  Think of how Alexa and other personal […]

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s


%d bloggers like this: