The high-tech fallacy in HR departments
Who fits the job? Algorithms are supposed to decide more and more often - supposedly completely unprejudiced, but according to "soft" criteria. Why not consult an oracle?
For some time now, strange messages have been circulating on the internet. There is talk of "robot recruiting". From algorithms that are supposed to distinguish which candidates are suitable and unsuitable as soon as the application is received and during the first interviews. And that supposedly free of any discrimination.
An insurance company provided a closer look. The computer conducts its own initial conversations there. But anyone who suspects that this is about examining applicants' financial expertise, important foreign language skills or even mathematical and logical thinking skills may be surprised. The robot has another mission. It is programmed to ask applicants about their preferred travel destinations, for example, or what leisure activities they like to pursue most. There are hardly any limits to the range of topics for a seemingly casual small talk.
It is easy to see that such an introductory interview cannot be used to scrutinize an applicant's professional qualifications. Literally it even says: "What the applicants can do professionally is of secondary importance. It also doesn't matter what they say, what matters is how they say it: based on voice, sentence structure, intonation and vocabulary, the insurer finds out whether a Applicant fits the company culture - or not." With this information, an implicit and far-reaching distinction is made: namely between the assessment of a specific qualification (the previous standard of performance assessment) and that of a diffuse fit (the introduced, rather "cloudy" innovation).
"Zeitdiagnosen" is a project in cooperation with the publisher Springer VS. In this column, changing experts from the social, media and political sciences comment on current debates of our time once a month.
The fact that an examination of choice of words and correctness of language is used to find out whether an applicant fits the company is perhaps something to be gained from. But where does it lead, does intonation, melody or dialect come into play? Such characteristics of a person usually reveal little of substance about whether someone is suitable for a position; they are also largely anchored in private life. However, the private sphere usually has little to offer for organizations that could also be used directly for professional activities. The attempt at a superficial personnel check is likely to be interpreted as encroaching. After all, it is precisely the hallmark of the modern world of work that people are not simply integrated into organizations "with skin and hair", but provide their workforce in functional roles.
Algorithms "liquefy" the agony of choice
The use of algorithms in human resources is primarily aimed at two goals: On the one hand, one would like to reduce costs and, on the other hand, avoid subjective influences when selecting personnel. In fact, algorithms can reduce the classic decision-making process, and not just when filling vacancies. They are also used for credit checks at banks, risk analysis procedures and forecasting of purchasing decisions and voting behaviour.
The question remains what actually happens when an algorithm decides? Strictly speaking, such a computer program does not make decisions in the way that is typically expected in an organization; but rather calculates a result. In other words, algorithms "liquefy" the agony of choice when making decisions; in such a way that they achieve results quickly, discreetly and without social confrontation. This should also explain why algorithms for human resources work currently appear so promising: they offer the promise of circumventing relatively complex interactions in personnel evaluation in a relatively "socially weak" manner. Precisely because HR is a demanding communicative setting, the simplicity of this computational solution seems so spicy.
What gets lost in this "desocialization" of personnel selection is the banal fact that the robots must of course continue to be loaded with data before they can even work. And here an almost paradoxical observation can be noted: people assist the machine in assisting people. The algorithms relieve people of discussion; In doing so, they delegitimize different assessments and perspectives by suspecting a lack of professionalism and forming uniform judgments. The applicant does not fit the culture, does he use the wrong words or is his voice irritating? – The algorithm, not a human, then finds the right answer, what humans have to do now. Even if the computer doesn't "think up" anything that it hasn't already "told" it, i.e. programmed it in.
Discriminate particularly effectively with corporate culture
One could therefore understand algorithms in personnel selection as procedures that inhibit discourse rather than promote it. First, they are based on assumptions that have literally been "preset" in them; secondly, they replace the tried and tested negotiation between people, the struggle for the decision. If algorithms, as in the example above, are to evaluate the fit to the corporate culture, it should be pointed out that hardly anything in organizations is as difficult to evaluate or even automate as their culture. Culture, i.e. the informal character of an organization, is not simply given, it is normative, also contains prejudices, even forms of subtle exclusion and disadvantage. To put it cynically: If you want to discriminate particularly effectively, you are most likely to use cultural criteria as a basis for the job interview.
All in all, the cultural fit-oriented algorithms in recruiting are reminiscent of a kind of oracle, a "digital Delphi". An innovation that basically does not deliver more than - measured by conservative performance assessment - invalid prophecies, just in high-tech format. The comparison to the long-popular "graphological method" also comes to mind. Applicants had to do handwriting exercises. All sorts of nonsense about personality was then read from these. The procedure disappeared from companies after psychological research exposed it as dubious. One might as well have entrusted the selection of personnel to a palm reader.
Maybe the modern palm reader is an algorithm? In any case, it is justified to approach algorithm-based recruiting with reservations. The standards of suitability diagnostics, i.e. the task-oriented examination of professional and personal suitability, have been elementary for the selection of personnel for decades. The professionalization of the HR department is also largely due to this; although every personnel process is not immune to subjective influences. It would therefore be in the interest of HR managers and recruiters to open a critical discussion about such technologies, with which the reputation of their work can possibly be increased far less than one might expect in all too naive fascination at the moment.
