Until now the algorithm as such has not yielded very accurate results when it comes to contract
personal. The most striking case was that of Amazon. In 2014, the company Jeff Bezos, in experimental way, began to use programs computer to search workers that will fit your needs. However, the results were disappointing, and the company had to abandon the project.
The problem is that the algorithm misinterpreted some data. Being the majority of Amazon employees of male sex (the typical engineers of Silicon Valley), the program concluded that the best thing for Amazon was to sign males. And it was eliminating the candidates from the selection women or harming them.
The algorithm discriminated against women and gender was not a criterion
The algorithm is not infallible, of course. Repeat certain common patterns (consider that if it is worth in most cases, it will work for the specific case) and, if you do not introduce the appropriate nuances, as in the case of Amazon, errors occur. It is not a problem as much of the algorithm as of how it is programmed. Experts explain that putting excessive variables generates "noise" and distorts the results, so what a company has to ask itself is what data is relevant to the search. The promotion of gender equality, in this case, it was.
Years have passed since then, and the most recent research is fine-tuning. "One of the classic paradigms is that of recommendation," explains Cecilio Angulo, director of the Ideai, a research center in artificial intelligence and data science that has just been released this week at the UPC. It's a bit what happens when we buy a book on the internet and the algorithm suggests similar books that we think we might like.
But this system has some shortcomings when it comes to examining professional talent. "We do not know, for example, if a company that needs personnel will have the same needs within five years," says Angulo, who works with Seat and cites the case of the automotive sector. "Maybe, with the arrival of the autonomous car, in the future we will have to prioritize other technological skills that are not taken into account now," he says.
Not only do companies look for workers, but also the opposite: the algorithm can also help a recent graduate to identify the companies that best fit their academic profile. "The match between demand and labor offer will increasingly resemble an application to find partners, "says Angulo, who knows the case of Esade students closely. "In this context, in a few years the curriculum vitae will disappear."
In effect, when a resume is presented, it can happen not so much to lie -which also-, but to exaggerate some strong points. If the algorithm is limited to examining the presumed virtues of the candidate, the result can be misleading. But if, on the other hand, the program is able to find out, then the precision will be greater.
Let's say that the algorithm manages to verify that the candidate has a high degree of English after examining their data in social networks or in the cloud (for example, discovers that a text of his circulates in English), then the chances of fitting will be more effective ( the last word to sign, in any case, will correspond to a human being).
In the future, artificial intelligence will seek to match companies and candidates seeking affinities
Also, if Amazon had problems with the genre years ago, in the future, companies and workers will have to worry about examining and leaving the computer traces of their respective trajectories. Maybe someone is looking for a company not only because they want to perform a certain task, but also to have the opportunity to grow and develop their creativity in this firm. It has, as Angulo calls it, "subjective interests". Well, if the company records that it does offer this possibility (for example, when Google says it leaves its employees free time to investigate other issues), then the algorithm will be able to identify it.
The reverse is also understood: if a newspaper is looking for a particularly creative journalist, the algorithm should be able to find articles in the network that contain words, references or some original structure to keep track of the author in question. And that's how the couple will be formed.