Basile Simon (blog) Data, hackery, stories

On recommendation in News

A couple of weeks ago, I had a chat with Paul Conge, from French left-wing newspaper Marianne about recommendation in news. His piece, These robots shaping the Internet to your image, is interesting but calls for some explanations - and a couple of words in English as well.

Important correction to Conge’s article: I do not work on content recommendation for the BBC News website. I took part in several workshops about this topic, organised by different departments, indies, and contractors, and I attended a bunch of sessions in which this topic was mentioned and discussed. But this is not what I do.

Content recommendation

Essentially, the gist of this topic is summarised by Conge’s first sentence:

“Sooner or later, there might be as many versions of the internet as there are users.”

Personalising the web is the contrary of serving identical webpages to all visitors; it is presenting different versions to different readers, according to a set of criteria.

The Three Ways

I outlined to Conge three ways of looking at content recommendation, from a publisher’s perspective:

  • Using a set of simple and basic data such as the time of day and the day of the week one visits a webpage, and the approximative geolocation based on one’s IP address, it is possible to build different types of news websites. For example, a publisher might want to have different home pages on the weekend and on a week day, as well as offering a different edition to an overseas and to a domestic reader. This is fairly simple.
  • Using more complex techniques such as profiling and machine learning, a publisher could leverage the use of cookies and track its users across its website to build a more comprehensive picture of who this particular person is. It might be worth presenting a user who reads only the “Sports” section and connects to the Live pages when big games are one with more sports-related news, for example. This tracking is based on a legalese and quite unsuitable view of what is consent on the web, and also leads to challenges regarding the data thus collected.
  • A publisher could also recommends to its readers to create a profile and fill in informations about themselves to build their profiles. These informations provided can be changed, edited, and removed at any time by the owner of the account.

Now, leaving aside the first solution which is easy to implement, choosing between the second and third ones is a trade-off.

The Two Dangers

The whole concept of recommendations and personal curation finds its sweet spot between the hammer and the anvil.

On the one hand, it is based on the assumption that a fully editorial-led point of view is not desirable, or at least not ideal. “You have to read this.” “You should know that.” Isn’t it a bit patronising?

On the other hand, there’s the history prison. Just because you followed avidly the General Election doesn’t mean you’re interested in the Labour leadership race. And just because you were into pop star celebrities at some point doesn’t mean that you should be served only this kind of content.

Give (control) to the user

The users know better. They know what they’re ready to share with you; what they want; what and when they want to change (something) about themselves.

Give this control to the readers. Ask them to tick boxes, to subscribe to topics, to create an account to take their personalised experience across their devices without using your evil cookies, to export and delete the data you have about them, to opt in and out of some of your services. They know better.

Readers come to your website because they respect your judgement and trust you to inform them - you can probably respect theirs.

Now, of course, there’s a limit, outlined by Lis Howell (whose article I don’t agree with in full):

Sometimes, I would contend, it does have to be one to everyone, not one to one. That is what news is. The issue of the personal curation of the news items which we want, but not the news items we need, is something which is an issue for all modern news providers. But currently only the BBC is paid by the public to get it right.

I agree. Public service, public interest, and our mission as journalists is also to push important things in front of the audience.

There are things everybody ought to know, after all.

Edit 21st August 2015:

Check out this article from Michael Skelton about the new BBC Homepage: Pick your own mix with customisation