<trp-post-container data-trp-post-id='25695'>Algorithms from recommendation or the question from the relevance

Yesterday, I was strolling through a local bookshop, one of those bookshops that have managed to withstand the combined onslaught of supermarkets, Amazon and Fnac.com, simply because the shop assistants read and love to read... and, above all, to share their discoveries.

Amazon suggests loads of books: the latest Goncourt because I bought last year's, the Femina because everyone who bought the last two Goncourts ordered the Femina - and a book on embroidery because I gave a book on crochet to my elderly aunt! I'm just messing about, even if the online bookseller's recommendation algorithm is a bit more sophisticated; in any case, as far as Amazon is concerned, you don't need to know any more than that: you've left your mark on the web, and other people even more so, so you should read this or that.

I'm still free to refuse... for now.

Many people buy the latest literary prizes, not to read them, but as gifts: it makes a nice present - and it's necessarily a good choice, since it's endorsed by a prestigious jury; but obviously, Amazon's algorithm doesn't take these subtleties into account.

In this local bookshop, small texts were posted on some of the books: the staff's reading notes.

Often, you enter these shops with a specific idea in mind - the latest Murakami, for example - and as you stroll around you discover unknown authors and works, you browse the little cards and let yourself be seduced... and you leave with your arms full, happy with your discoveries.

And Amazon will never be able to offer us this pleasure: recommendations that don't come from our past, but from the passion of real readers.

And yet, recommendation algorithms are beginning to make their mark in a wide range of sectors, from dating with Tinder to human resources with a host of start-ups such as Hunteed, Kudoz and Clustree - the latter of which managed to raise €2.5 million last year!

There's no longer any need for dozens of long and tedious job interviews: Clustree looks to see if you already have the ideal candidate in-house, while Kudoz gently taps into LinkedIn to see if anyone matches your expectations.

And they'll deliver the perfect collaborator, like Tinder delivers a one-night stand and Amazon a novel for your next holiday.

Of course, just as with Amazon we are depriving ourselves of great books of which we knew nothing, even the existence of the author, so here we will necessarily miss out on the disruptive candidate who could give his company a new lease of life: recommendation algorithms protect us from taking risks that are sometimes salutary.

François laurent

They we have fact confidence. Discover our achievements