Crowdfunded OpenSCHUFA Project Wants To Reverse-Engineer Germany's Main Credit-Scoring Algorithm

from the opening-the-black-boxes dept

We've just written about calls for a key legal communications system to be open-sourced as a way of re-building confidence in a project that has been plagued by problems. In many ways, it's surprising that these moves aren't more common. Without transparency, there can be little trust that a system is working as claimed. In the past this was just about software, but today there's another aspect to the problem. As well as the code itself, there are the increasingly-complex algorithms, which the software implements. There is a growing realization that algorithms are ruling important parts of our lives without any public knowledge of how they work or make decisions about us. In Germany, for example, one of the most important algorithms determines a person's SCHUFA credit rating: the name comes from an abbreviation of its German "Schutzorganisation für Allgemeine Kreditsicherung", which means "Protection Agency for General Credit Security". As a site called Algorithm Watch explains:

SCHUFA holds data on round about 70 million people in Germany. That's nearly everyone in the country aged 18 or older. According to SCHUFA, nearly one in ten of these people living in Germany (some 7 million people) have negative entries in their record. That's a lot!

SCHUFA gets its data from some 9,000 partners, such as banks and telecommunication companies. Incredibly, SCHUFA doesn't believe it has a responsibility to check the accuracy of data it receives from its partners.

In addition, the algorithm used by SCHUFA to calculate credit scores is protected as a trade secret so no one knows how the algorithm works and whether there are errors or injustices built into the model or the software.

So basically, if you are an adult living in Germany, it's a good chance your financial life is affected by a credit score produced by a multimillion euro private company using an automatic process that they do not have to explain and an algorithm based on data that nobody checks for inaccuracies.

A new crowd-sourced project called OpenSCHUFA aims to change that. It's being run by Algorithm Watch and Open Knowledge Foundation Germany (full disclosure: I am an unpaid member of the Open Knowledge International Advisory Council). As well as asking people for monetary support, OpenSCHUFA wants German citizens to request a copy of their credit record, which they can obtain free of charge from SCHUFA. People can then send the main results -- not the full record, and with identifiers removed -- to OpenSCHUFA. The project will use the data to try to understand what real-life variables produce good and bad credit scores when fed into the SCHUFA system. Ultimately, the hope is that it will be possible to model, perhaps even reverse-engineer, the underlying algorithm.

This is an important attempt to pry open one of the major black boxes that are starting to rule our lives. Whether or not it manages to understand the SCHUFA algorithm, the exercise will provide useful experience for other projects to build on in the future. And if you are wondering whether it's worth expending all this money and effort, look no further than SCHUFA's response to the initiative, reported here by netzpolitik.org (original in German):

SCHUFA considers the project as clearly directed against the overarching interests of the economy, society and the world of business in Germany.

The fact that SCHUFA apparently doesn't want people to know how its algorithm works is a pretty good reason for trying to find out.

Follow me @glynmoody on Twitter or identi.ca, and +glynmoody on Google+


Reader Comments

Subscribe: RSS

View by: Time | Thread


  • icon
    Roger Strong (profile), 19 Mar 2018 @ 10:40pm

    Reputation Gatekeepers

    The Information Age started without the Industrial Age ending. While the Information Age hasn't ended, I've seen a few comments here and there to the effect that we've now entered the Reputation Age.

    The era of "fake news" made it necessary. Now the reputation of the news source is as important as the news. Credible citations are required.

    But it's not just debunking Trump's "alternative facts", climate change denial or Alex Jones' claims. Or the false claims about SESTA and Net Neutrality. Or the New York Times vs. Breitbart.

    This story is about being a gatekeeper not for information, but for reputation. Credit scores are about reputation, and SCHUFA is the gatekeeper. OpenSCHUFA wants to change that.

    Scientific journals are about reputation. There are many deservedly poor-reputation journals, but there are some that have cultivated a reputation so good that merely being published in them establishes a scientist's reputation. And so Elsevier becomes a gatekeeper not just for papers, but for scientists' reputations.

    Would-be reputation gatekeepers have had some big misses. Reputation Management Consultants keep encountering the Streisand Effect and now the occasional angry judge. Facebook and Twitter have accidently taught everyone that a large number of followers does not equal a good reputation.

    But they're experimenting and learning. Politicians depend on reputation, so it'll be easy to influence them to pass legislation helping to create gatekeepers. Its early days in the new age.

    reply to this | link to this | view in chronology ]

    • identicon
      Anonymous Coward, 21 Mar 2018 @ 12:32am

      Re: Reputation Gatekeepers

      Now * the reputation of the news source is as important as the news.

      Always was - what is said is secondary to who said it, because the latter allows you to infer why it was said, and this in turn shows a bigger picture. Hell, even here - there are people like PaulT, ACs and outOfTheBlue, and boy, are they treated differently.

      *Emphasis mine.

      reply to this | link to this | view in chronology ]

  • identicon
    Anonymous Coward, 20 Mar 2018 @ 12:18am

    I'll bet they find a surprise

    I would say that there are better than even odds that the OpenSCHUFA project analysts will be scratching their heads due to unexplained variance after performing the best fit linear and nonlinear regressions on all the observed variables. And they will come to find that this variance represents undisclosed - and possibly illegal - data used in calculating the SCHUFA score. For example, information that comes from breaches of privacy or government records like medical history, employment records, ethnicity, religion, arrests, tickets, and tax information.

    That will be a very interesting can of worms when opened, because Europeans value their privacy much more highly than us Americans.

    reply to this | link to this | view in chronology ]

    • identicon
      Anonymous Coward, 20 Mar 2018 @ 2:52am

      Re: I'll bet they find a surprise

      This is my suspicion, and I suspect that it affects lives on a variety of discriminatory scales, but that's just an intuition.

      reply to this | link to this | view in chronology ]

    • identicon
      Anonymous Coward, 21 Mar 2018 @ 12:37am

      Re: I'll bet they find a surprise

      information that comes from breaches of privacy or government records like medical history, employment records, ethnicity, religion, arrests, tickets, and tax information.

      Highly unlikely. If it was true, the trainwreck that the SHUFA would create would be enormous. What was it - fine of 4% of yearly revenue?

      reply to this | link to this | view in chronology ]

  • This comment has been flagged by the community. Click here to show it
    identicon
    Saim Locus, 20 Mar 2018 @ 12:31am

    Microsoft Support

    the Scoring algorithms for engineers are very I useful just like I saw that but we are the service provider of the Microsoft if you have any issue related to the Microsoft then you can visit us because we have all types of information.

    reply to this | link to this | view in chronology ]

  • icon
    Andrew (profile), 20 Mar 2018 @ 7:11am

    Let's use VI

    :1,$s/Germany/United States/g

    reply to this | link to this | view in chronology ]

  • identicon
    Anonymous Coward, 20 Mar 2018 @ 7:23am

    Credit Rating ... the new jim crow

    reply to this | link to this | view in chronology ]

  • identicon
    Châu, 23 Mar 2018 @ 12:56pm

    AI

    Good project for use AI research and algorithms.

    reply to this | link to this | view in chronology ]


Add Your Comment

Have a Techdirt Account? Sign in now. Want one? Register here
Get Techdirt’s Daily Email
Use markdown for basic formatting. HTML is no longer supported.
  Save me a cookie
Close
Have a Techdirt Account? Sign in now. Want one? Register here
Get Techdirt’s Daily Email
Use markdown for basic formatting. HTML is no longer supported.
  Save me a cookie
Follow Techdirt
Techdirt Gear
Show Now: Takedown
Advertisement
Report this ad  |  Hide Techdirt ads
Essential Reading
Techdirt Deals
Report this ad  |  Hide Techdirt ads
Techdirt Insider Chat
Advertisement
Report this ad  |  Hide Techdirt ads
Recent Stories
Advertisement
Report this ad  |  Hide Techdirt ads

Close

Email This

This feature is only available to registered users. Register or sign in to use it.