According to Sarah Bana, a researcher in artificial intelligence at Stanford University, it is possible to use an AI to analyze the words contained in a job offer and determine with precision the salary proposed by the employer.
Advances in machine learning and artificial intelligence have reached new heights in recent years. At the beginning of June 2022, a news incident made the headlines in the specialized media: a Google engineer claimed that his AI was endowed with human emotions, the machine considering itself to be a sensitive person.
Among the other works of Google in the field, we also note the presentation of Imagen, an AI capable of creating complex images from a simple description. However, Sarah Bana, doctor holder of a research grant at the Digital Economy Lake of the prestigious university of Stanford, has just exploited AI in an unexpected field.
To put it simply, it ensures that it is possible to use artificial intelligence to analyze words in job postingsand accurately predict salaries to which candidates may apply. “It turns out that we can use the text of job advertisements to evaluate the salary characteristics of jobs in near real time. This information could make job applications more transparent and improve our approach to workforce education and training.” she assures.
Also read: Google Maps – an AI blocked 100 million fraudulent changes in 2021
This AI can tell you how much you will earn
To conduct these tests, Sarah Hana accessed more than a million pre-pandemic job postings from Greenwich. HR, a platform that brings together millions of jobs published online. Then she used Google BERT (Bidirectional Encoder Representations from Transformers), a language model developed by the firm in 2018 that has significantly improved performance in automatic language processing.
Using BERT, she trained an NLP model on the text contained in over 800,000 offers. Once the model had worked out, she tried it on the remaining 200,000 offers. Turns out he made it predict the salaries associated with the various offers with an accuracy of 87%. By simply analyzing the job title and the geographical location, this percentage still reached 69%.
In her next work, the specialist intends characterize the contribution of various words to the prediction of wages. “Ideally, we’ll color the job ad words red to green, with dark red words being linked to lower pay and dark green words being linked to higher pay.” she explains.
Source: Ventura Beat