In other words, based on your consumption of news, your search patterns, and a deep analysis (semantic, tonality, implied emotions) of your mail and your posts — matched against hundreds of millions of others — Google will be able to suggest a link to the profile of an artist in Harper’s when you dropped in Google News to check on Syria.
Frédéric Filloux, News: Personalized or Serendipitous?, Monday Note.
Filloux interviews Richard Gingras (senior director of news and social products at Google) about how Google is planning to serve content to readers that is both personalized and serendipitous:
According to him, “Today’s news personalization is very unsophisticated. We look at your news reading patterns, we determine that you looked at five stories about the Arab Spring and we deduct you might like articles about Egypt. This is not how it should work. In fact, you might be interested in many other things such as the fall from grace of dictators, generation-driven revolutions, etc. These requires understanding concepts”. And that’s a matter Google is working on, he says. Not only for news, but for products such as Google Now which is the main application of Google’s efforts on predictive search.
How it’ll actually pan out, we don’t know, but there are a couple of questions that come out of this that are worth keeping in mind:
01. How do recommendation engines (currently) work?
Recommendation engines, otherwise known as recommender systems, suggest content based on previous behavior or purchases. Such systems typically use one of two approaches: Collaborative filtering creates a predictive model based on a user’s previous interactions such as products purchased or viewed. Content-based filtering looks at content or item characteristics and suggests content with similar elements. Amazon, Netflix and music services including Pandora and last.fm use recommendation engines.
02. Why preserve serendipity?
Four years ago, Mathew Ingram wrote about the serendipity defense—one of the main arguments for a bundled news package: basically, newspapers let you easily stumble upon fascinating articles that you’d never think to look for by yourself. The advantage that newspapers have over the internet on this is that once you pick one, you can have a good sense of what quality to expect from it. You know the brand, you can let your critical guard down and the whole experience is fun and easy. While the internet might be the ultimate serendipity machine, achieving the same results requires much more: know-how, patience, a critical eye on the trustworthiness of the source, and regular exploration.
03. Why prescribe content?
While the web allows us to find endless amounts content, that content often ends up being finer and finer bits of the candy we love, which encourages us to exist in echo chambers of information-decadence. Those who believe that news services should deliberately prescribe content to help readers diversify the perspectives they are exposed to have yet to find a way to do so that doesn’t feel like an eat-your-vegetables approach. Plus, prescribing content to readers for the benefit of a “balanced” news diet would likely require too many assumptions about value to ever be widely accepted.
FJP: If you’re a proactive news reader, you can create your own diet of publications or people both in and outside your comfort zone and maintain a personalized, serendipitous, diverse set of sources. If you’re not, you’ll rely on recommendation engines to do the work for you. For the creators of those engines, it’s a tricky mix to find the right balance between serendipity and prescription. And unfortunately, many of the concocters of the magic potion are for-profit companies that skew the formula for their own benefit.—Jihii
Instead of seeking an engaged audience — that’s a metric better suited for movies and prime-time TV — we in news should be seeking an informed public, using new tools to make them better informed with greater relevance and more efficiency. Instead of measuring our success by how much more time we can get them to spend with us, we should measure it by how much less time they need to spend with us to reach their own goals.
Jeff Jarvis, Maybe News is Just More Efficient, BuzzMachine.
FJP: Jarvis goes on to discuss a hypothetical news service that accomplishes this task of efficiency by serving the public news that uses a (very) smart algorithm to bring each reader a hyper-personalized news stream. An obvious issue with this is selection bias, and the possibility that one will end up consuming a very narrow slice of the perspective pie (read about the perils of algorithmic curation here). Also see the comments below Jarvis’s article for some interesting points made by readers.
A big part of the reason he is such an effective aggregator for both audiences and news sites is that he actually acts like one. Behemoth aggregators like Yahoo News and The Huffington Post have become more like fun houses that are easy to get into and tough to get out of. Most of the time, the summary of an article is all people want, and surfers don’t bother to click on the link. But on The Drudge Report, there is just a delicious but bare-bones headline, there for the clicking. It’s the opposite of sticky, which means his links actually kick up significant traffic for other sites.