University degree: When an AI evaluates the university grades

Icon: enlarge

AI now reaches into almost all areas of life - including recruiting

Photo: Henrik Sorensen / Digital Vision / Getty Images

Expensive suits, thick watches, a lot of trips, well, as far as Corona allows, plus a starting salary beyond the 50,000 euro equator - many students dream of joining as a management consultant straight away after university.

People like Stephan Butscher can tell from the level of their email inbox: He is Chief Human Resources Officer at the consulting firm Simon Kucher & Partners with 1,400 employees worldwide.

Every year he receives several thousand applications, says Butscher, far too many for too few positions.

Basically, graduates of many courses of study would be an option for him, "basically everyone in which business or analytical work occurs somewhere": business administration, economics, industrial engineering or business informatics - but natural scientists and psychologists are now increasingly being used.

Difficult to keep track of things: who is how good, how well-known is the university in the field of study?

Is an average grade of 2.4 at a particularly good university better than a 1 at a B-Uni?

Or is it the other way around?

"With many thousands of different courses of study, it is impossible for a person to evaluate all the important contextual information with any degree of objectivity."

Philipp Seegers, founder of CASE

That is why Butscher has been working with the so-called CASE score for three years.

A sober percentage that shows how the graduate fares in comparison to the competition: A CASE score of 14 means that the applicant belongs to the top 14 percent of her year in her subject, across Germany.

So: the lower the better.

The CASE-Score was developed by the Bonn-based start-up Candidate Select.

Information on the distribution of grades in the various courses of study at the individual German universities, dating back to 2004, is used as the data basis Taken personality tests.

An algorithm leaves out emotions

"With many thousands of different courses of study, it is impossible for a person to evaluate all the important contextual information such as grades or quality with any degree of objectivity," explains Philipp Seegers.

He founded CASE in 2016 together with his partner Jan Bergerhoff.

When analyzing the data, the two therefore rely on algorithms and artificial intelligence.

»Our decisions are shaped by emotions that can fluctuate.

An algorithm always does the same thing.

This makes the results better, but also reproducible and fair, ”says Seegers.

In concrete terms, this means: algorithms recognize and sort universities, subjects and degrees;

In addition, an AI optimizes the CASE scoring.

"The aim is that the result of our score should be constant on average," explains Seegers.

It is not universities that are included in a ranking, but specific study programs.

More from »The Boss Machine«

Icon: Spiegel PlusIcon: Spiegel PlusBusiness psychologist on job interviews: Are intelligence tests the better job interviews? An interview by Jens Radü

An example: A business administration student with a 2.0 bachelor's degree at the University of Rostock was among the top ten percent of her year in 2013.

Quite different at the University of Konstanz: With a master’s degree of 2.0 in 2016, the example student would be clearly below average;

73 percent of the students there had a better grade.

The CASE score of a 2.0 in the Rostock Bachelor is significantly better than that of a 2.0 in the Konstanz Master.

Sounds complicated?

Absolutely.

So try it out for yourself: In the following infographic, for which CASE provided the data, you can compare business administration courses at various German universities.

Accepted final grade: 2.0.

In the lower third, the CASE score is displayed depending on the course and graduation year, embedded in the overall distribution.

Let's go:

Is that fair now?

Devaluing 1.0 in subject X just because University Y grades less strictly according to the data?

What can the student do about it?

"This is all very sensitive and we are of course aware of that," explains founder Seegers.

That is why it is particularly important to him and his team to collect clean data that does not mislead the system just because one table says "studies" and the other maybe "college education", but the same thing.

The AI ​​is just one tool among many - but it is indispensable when it comes to millions of bits of information that need to be processed.

"In many applications, AI models represent a real revolution because much better predictions are possible."

Icon: enlarge

Katharina Zweig

Photo: Felix Schmitt

But there is also the risk of overfitting: The model fits perfectly - but only for a certain sample of degrees.

But only in individual cases, not in general.

"There is no evidence that technology can manage to extract only the good and right from allegedly prejudiced human decisions."

Katharina Zweig, head of the Algorithm Accountability Lab

Bioinformatician Katharina Zweig, who heads the Algorithm Accountability Lab at the Technical University of Kaiserslautern and researches the traceability and regulation of artificial intelligence, also warns of excessive trust in artificial intelligence: “There is no evidence that the technology can create it from the Allegedly prejudiced human decisions only pull the good thing out of the right thing. ”She therefore calls for more independent evaluations of these systems before they are used.

Because AI is spreading more and more functions in more and more industries.

CASE-Score helps with the selection

After all, when asked whether the CASE score actually measures what it should - namely the quality of qualifications and thus also that of the training of graduates - Stephan Butscher from Simon Kucher & Partners has a short answer: yes.

"The CASE score is actually a good indicator for us to find applicants who are a good fit for us," says the HR manager.

Candidates with a poor CASE score were invited to job interviews in an extensive pilot phase - but they received an offer significantly less often after the interviews.

And the careers of the junior consultants, who were hired on the basis of their good CASE scores, have progressed significantly better over the years: from consultant to senior consultant to manager, and that at a rapid pace.

»Employees fresh from university are promoted once a year on average.

For top performers, the value is closer to 0.8 years.

And there are a particularly large number of these top performers who also have a very high CASE score, ”reports Butscher.

However, the CASE score is of course not the only criterion that matters.

And in his industry in particular, Butscher cannot imagine that an AI will soon be directing the entire hiring process: »We are looking for top people with a wide range of skills and qualities.

Without personal contact and personal conversations, even quite intensive ones, we will not get by. "

Icon: The mirror