′Preference aggregation algorithms′ know exactly what you like | Science| In-depth reporting on science and technology | DW | 22.08.2011
  1. Inhalt
  2. Navigation
  3. Weitere Inhalte
  4. Metanavigation
  5. Suche
  6. Choose from 30 Languages


'Preference aggregation algorithms' know exactly what you like

In this fast-moving technological age, websites and applications are constantly making us offers based on what they think we might like. And they're often spot on. So how do they do it?

Silhouettes of two people at laptops. Facebook logo above them

Facebook knows what you like and is happy to share its information

Last Friday, the privacy commissioner for the northern German state of Schleswig-Holstein said that he would go after private and public sites featuring Facebook's "like" button, which allows web users to rate the websites they encounter.

The local data protection authority worries that Facebook is collecting too much information each time someone clicks a simple "Like" button.

The agency could slam websites operating in Schleswig-Holstein with up to 50,000 euros ($72,000) if they continue to use the "Like" button or manage fan pages on Facebook's website after the end of September.

Experts say this is just the latest example of how Germany is balancing privacy concerns with research that tries to tailor new social media products, services and smartphones apps like never before.

An increasing number of tech startups know when we rate something we like on Facebook or Google, and use what are known as "preference aggregation algorithms" to build a comprehensive profile of our tastes.

Different kinds of chocolate

Can't decide? Ask a preference aggregation algorithm for help

Research that was once confined to obscure areas of mathematics and information science is now influencing the tech sector in a big way.

Alexander Kröller, a computer scientist at the Technical University in Braunschweig is currently working on a project which required him to create a preference aggregation algorithm to find out about children's preferences for certain foods.

"You can't go to a three-year-old and say here are eight pictures of fruit, please put them into the order in which you like it, because that's not how kids work," Kröller told Deutsche Welle.

Accounting for inconsistency

Instead, he presents the children with an automatic system which shows two food items at a time, and asks them to click the one they like best.

Sounds simple, but kids, like adults, are anything but consistent, so there is not always a clear-cut ranking that computers can understand. Kröller has to factor this into his system.

"What we are working on is to construct algorithms that can work with this inconsistent data," Kröller continued.

The social graph

Amazon.de logo

Amazon frequently suggests products it thinks customers might like

And these algorithms, or others like them, are widely used by social media startups like Gidsy, a new Berlin startup, which is set to launch in September.

Gidsy hopes to enter the "experience" market, helping people who want to host pop-up restaurants or shops, run a walking tour or teach a beginner's language class to contact those interested in pursuing such activities.

Facebook maintains its own set of data showing how various people and their interests are connected: a concept it calls the Social Graph. Edial Dekker, Gidsy's founder, argued that this data is invaluable to the service his company will soon be offering.

"For us it's very important that we know our users and know the people that are organizing things," he said. "We know who your friends are and where they're from and we try to make it relevant."

Location, location, location


Getting smarter all the time

His company isn't the only one jumping on the preference aggregation algorithm bandwagon. Earlier this month, another Berlin start-up, EyeEm, began trading.

The site combines social networking with photos taken on Smartphones. As the company's founder Florian Meissner explains, the idea is to add the photo's location to the algorithm, and consequently suggest networks that might be interested in that location.

"Location is the basic assumption that we base our intelligent suggestions on. We try to blend it in with an activity layer," he said. "When you think about hanging out [after] a conference, we know you're having a meeting or an after-work beer with your colleagues."

It may just be only a matter of time before smartphones know all of our favorite tipples and can even order one in advance.

Reporter: Jonathan Gifford, Berlin / tkw
Editor: Cyrus Farivar

DW recommends