While researchers have been interested for several years in the field of personality detection on the internet, Frédéric Piedbœuf is the first to demonstrate that this is also possible on a professional platform like LinkedIn. The doctoral student at the Université de Montreal (UdeM) has developed an algorithmic method capable of detecting certain character traits in users of this social media.
“Response rates are often quite low when human resources contact people on LinkedIn,” remarks Frédéric Piedbœuf. How then can communication be improved between recruiters and potential candidates found on social media? “By personalizing the messages,” he replies. This is where his algorithm comes into play: “It will allow you to know more about the candidate before you even contact them, to then personalize the approach and get a better response rate.”
This intelligent tool therefore represents a saving of time, and also of accuracy. The algorithms analyze a large number of parameters more quickly and then combine them in order to define one or more character traits. It is work that humans do in a more instinctive and less objective way.
It was already possible on Facebook, Twitter and Instagram. That said, successfully extracting human characteristics from LinkedIn profiles presented an additional challenge. “It’s a very different social environment than Facebook, because people share more professional than personal information while nurturing a more polished appearance,” says Frédéric Piedbœuf.
Based on data from some 1,200 public accounts, the student trained his algorithm to identify certain personality traits and make connections in two categories: textual and non-textual information. In the first category, adjectives and word types used in biographies and texts published on the platform are analyzed. Each word will then be associated with a subcategory. For example, people who use words relating to vulnerability, social relationships, submission to authority or dependence on others will be associated with a conscientious nature.
Non-textual information is defined according to the number of friends, the number of companies where the person has worked during their career or their field. “Everything you do on social media gives some indication of personality, even not being active there every day,” said the 25-year-old researcher.
Thus, a person who has many friends, who posts very regularly on their newsfeed, who uses few punctuation marks and many adjectives will be considered an extrovert. Conversely, an introverted person, according to a study from the University of Sheffield, will tend to use punctuation marks more.
Currently, recruiters cannot use the algorithm in question because it is biased. “I have demonstrated that it was possible to detect personality on LinkedIn, but I selected users who had previously published their results on personality tests,” explains the student. “These are people who are very active on the network and are therefore not representative of the general population.”
It will therefore be necessary to carry out a new study on an unbiased population for the algorithm to be functional and accessible to recruiters, while ensuring that users’ privacy is protected.