ADAPTED FROM THE NEWSTORY BY YVES EUDES “Comment notre ordinateur nous manipule” published in LE MONDE on Saturday April 12^th 2014 , p. 1 and in the newspaper supplement “Culture et Idées” an investigation by Yves Eudes)

Recording information on the internet population starts off as soon as they sign up, with the help of number-crunchers and cookies, short coding lines installed on the browser by the websites they can surf. Specialized agencies will thus draw up a very accurate profile of Mr Average internaut who will be monitored whatever his whereabouts may be on the web. Here are only the premises of what is in store for the internauts since predictive analysis will fine-tune these techniques, thanks to AI software and math experts.

Little wonder if targeted advertising is displayed on your screen.There are cookies lurking in your computer. They came into play as early as 1994, the year when the web provided public access, and have gained momentum since then to the point that online advertising shot up to a $102 bn turnover in 2013. What‘s the upshot of all this? They are dotted on the websites by specialized companies which collect information, classify, analyze and aggregate it. Thereafter this information is sold. What information do they store? Your passwords, your shopping baskets, the pace of your online surfing, slow or swift, systematic or intermittent, thereby customized files are kept on databases. The advertising pitch will be displayed in a timely way and in the right format. Any alternative for you? Remove the cookies but when you navigate again other cookies will take over and if they are blocked the visited sites are most unlikely to operate. And cookies die hard. Those installed by Amazon are designed to last until 2037.

Want another case in point? Hold on to your hat. According to the study carried out by Yves Eudes, if you surf PriceMinister, the home page of the merchant site will greet you while 44 cookies issued by 4 specialized agencies will sneak into your browser. Should your interest focus on mobile telephony, 22 new cookies will remain such as barnacles clinging to the vessel of your curiosity. Are you lured into getting a Samsung? This time 42 cookies will join the cruise. Three clicks and you will be filed 108 times by forty or so databases. Suppose you had only half a mind to place an order for this smartphone and give it up, all of a sudden? OK, but Critéo, a French start-up, will be on stalk. It specializes in hooking second-hand prospects so further advertising will follow you on the heel, promoting the same product for days on end.

Critéo has a business model of its own. Outfitted with number-crunchers it will perform various operations in next to no time: spot the customer, inform the platform in charge of advertising space, make a bid, carry out the electronic negotiation and display the banner. It is in a make it or break it situation as in thirteen hundredth seconds, the game is over. If it is beaten, another competitor will take over the advertising space.

Admen are not at their wits’ ends with all this cutting-edge technology as the information revealed by cookies is crossed and cross-referenced again with other data items: your internet protocol address (IP for the aficionados) that identifies and locates your computer, the language used, your queries on search engines, your computer and browser model etc. So much the worse for the nominative information left when you filled out an information sheet or purchased a product. Every piece of information will be harnessed.Is this where the targeting ends ?

No way. All the particulars reaped from the real world (i.e. bank statements, credit cards, sales receipts …) can be matched in turn with that of the virtual world.

Acxiom company, specialized in this twin cross-referencing, sells profiles based upon 150 criteria. Amongst them, some cannot but puzzle you: “accommodate a dependent”, “have a cat”, or “like sewing”. Furthermore, your revenues, tastes, needs, gender, age, occupation, leisure activities, ethnic origin, zip code, medical history, family status, address, your car model, religion, your holiday destinations are all too well-known. Each 1000-client file fetches around € 0.60. Rates vary and soar to € 250 when it comes to a pharmaceutical company. Thus, the specialized agency can draw up a list of patients suffering from obesity (whose names are undisclosed, the term coined is “anonymized data) who purchase slimming drugs.

Let’s now try to put an end to some clichés, those purporting for instance that online products are less expensive across the board. You bet. If you browse a shopbot and then move on to a holiday site you will notice that the operator lowers the rates to align with the competitors’. If you come across a tour operator flaunting bargain prices for your next holiday destination but wish, out of caution, to be aware of other possible attractive offers, you will realize, to your amazement, by coming back to the initial site that the indicated price is now higher and the tantalizing promotion has vanished in the meantime. Or if you have availed yourself of a sleek, new-fangled machine, the site is likely to post a more expensive hotel room. Want a ploy to outfox them? Use another machine, which is by definition equipped with another IP address.

The latest trend hinges on predictive analysis: gone are the days when admen merely responded to internauts’ expectations and behavior, they now intend to forecast and anticipate their responses. A new discipline, which is still in its infancy, has entered into play: machine learning, a branch of AI. According to Franck Le Ouay, scientific manager at Critéo: “machine learning designates the ability of a program to adapt to a new situation”. Self-regulating and self-teaching algorithms are under development and these experiments are now galore. Erick Alphonse, a mathematician at Paris XIII University is developing the so-called Predictive Mix system. It will single out the internauts that make no purchase if no advertising is displayed but are poised to do so provided it is. They are said to be of the “receptive kind”. The computer collects a sample of the “receptive profiles” matching the population in a database and the advertiser will just have to target this particular group, thus saving money (since less advertising space is paid for) and achieving a higher return on investment.

Weborama, another French company, is making use of “the web of words”: tabling on a collection of written materials to be found in sites and forums, its linguists draw up a list of 6000 relevant terms and its mathematicians regard the web as a metric space where they study the distance between the words and if the latter are associated within the same sentence. Then, these words are grouped together in 177 themes such as insurance, gambling, sports etc. Alain Lévy, Weborama’s CEO, puts it this way: “this taxonomy has become our vision of the web. The reference is no longer the site but the word”. Further to agreements with agencies, Weborama installs cookies on millions of browsers, then keeps track of them on the web and collects the words used on the visited sites. To take up Alain Lévy’s terms, each profile is assigned with a matching cloud of words. Number- crunchers will project this cloud onto the database storing the sets of words and assign to each profile a grade per category.

By correlating the grades, for example 13 out of a maximum of 14 for words associated with fashion, 12 for those dealing with design but only 2 for sports and 1 for cars the web will infer who hides behind a cookie. It will be, say, a 34 to 49 year-old woman, passionate about fashion but indifferent to sport. The profile will be devoid of interest to some advertisers but will attract others. L’Oréal will be ready to pay 2 euros to have a banner displayed on her screen. Currently, Weborama has acquired 62 million profiles for the whole of France and each click is followed with new calculations and regular updates.

On its part, Dataiku has devised a software package that will enable executives with no cutting-edge training in IT to manage databases and launch into predictive analysis. Florian Donetteau, at the head of Dataiku, remarks that one has to admit that computers are capable of correlations unheard of so far and of findings equal or superior to man’s, in record time.

In a coming paper, we will examine why, although the customer’s own free will may be brought into question , it may not be over. According to several thinkers, such as William Bates, philosopher and Professor of the History of Technology at Berkeley University (California), there seems to emerge rising awareness …