AI promised to speed up the processes of staff recruitmentbut after a period of intensive use of AI by both companies and candidates, it has been shown that It’s more broken than ever.
Further proof of this degradation are cover letters which, although before the arrival of AI models were a clear differentiating factor, are currently worthless, as a study by Princeton University and Dartmouth College has shown.
Cover letters made a difference. The study ‘Making Talk Cheap: generative AI and Labor Market Signaling‘ carried out by Princeton researchers analyzed more than 2.7 million proposals on the Freelancer.com platform before and after the implementation of the LLM text generation models to create these cover letters.
Their conclusion is that, before using AI, attach a well-written and to show interest and knowledge of the position and the company to which one was applying, considerably increased the hiring options because the recruiters perceived that this was a very capable candidate.
Now they are wet paper. However, as the use of AI tools to generate these cover letters has spread, the appreciation of quality has improved so that candidates in the top 20% of writing skills were 19% less likely to be hired, while those in the lowest 20% increased their chances by 14%.
In other words, employers stopped associating a well-written letter with a competent candidate. This has meant that the differentiating factor that a well-written cover letter previously provided has disappeared, reducing the curve of possibilities between the best-trained candidates and those who are not so well-trained.
Letters submitted before the LLM models had a better chance of being hired than those post-LLM
AI makes hiring more difficult. The effect observed in cover letters has been extended to other areas of personnel selection, since AI distorts real capabilities of the candidates. It is true that its use increases the perception of quality of the candidates, but as the average quality of the group increased, companies began to trust less in the information provided by the applications.
He study ‘Does AI devalue communication? Theory and evidence of entrepreneurship and contracting at a global level’ carried out by researchers at Columbia University and Yeshiva, found a similar pattern in selection and entrepreneurship processes: access to AI reduced the accuracy with which recruiters identified the best profiles to fill a given vacancy by between 4% and 9%.
If everything is good, nothing is good. For decades, a letter well tailored to the offer served as proof of interest and commitment on the part of candidates. In labor economics, this is known as “signalling”: the candidate conveys their effort through the quality of the text. Generative models have thrown that signal to the ground.
The meta-analysis ‘The role of artificial intelligence in personnel selection’ concluded that the automation of selection processes with AI is eroding the traditional signals of merit that were transmitted through cover letters, emails or applications received by the hiring and human resources departments. In that sense, while it is true that AI has democratized competition in the job search, it has also made genuine talent less visible.
Who is behind the algorithm? The current degradation of those “clues” that allowed recruiters to locate the best talent, forces us to look for new ways to evaluate candidates. As and as they pointed From the technological employment platform Manfred, the use of AI has multiplied the number of applications, but the perceived quality has not improved at the same pace.
For this reason, many companies are choosing to implement more practical tests and face-to-face interviews in their selection processes. That is, eliminate from the equation the presence of AI for the last stage of the selection process. The unknown of this practice is knowing how much talent has succumbed to AI resume filtering prior to that first face-to-face interview.
In Xataka | Jeff Bezos assures that there is a type of employee who can never be replaced by an AI: inventors
Image | Unsplash (Vitaly Gariev)
GIPHY App Key not set. Please check settings