Detect burnout using natural language processing (NLP)

Occupational fatigue or burnout is a state of deep fatigue, both physical and mental, resulting from chronic stress at work. Masha Kurpiz-Bricki, a professor of data engineering at Biel’s Bernice University of Applied Sciences, and her team of scientists, backed by the Swiss National Science Foundation (SNSF), have just developed a method for identifying it based on automated text analysis. Research Title ” BurnoutEnsemble: Increased intelligence to detect signs of burnout in clinical psychology. In “Borders on Big Data”.

A survey of about 10,000 working adults in 8 European countries found that 18% of respondents felt stressed on a daily basis and 30% of them felt so stressed that they considered changing jobs. This stress can lead to professional fatigue or burn-out.

The clinical diagnosis of this syndrome is difficult because the symptoms are similar to those of depression or anxiety in others. To identify this, psychologists rely on psychological tests where patients fill out a questionnaire with a tick on a multiple-choice question, such as “Maslach Burnout Inventory”, which consists of 22 multiple-choice questions divided into three sections: burnout, depersonalization / loss of empathy. And personal performance evaluation.

An example of a simple question-answer: “I feel tired at the end of my work day: never / never / every day”. Some people do not dare to tick the “never” and “every day” answers or are tempted to lie to influence the results.

More complete questionnaires consisting of open questions can also be used to identify burnouts. Although they provide more relevant information, they require significant analytical work and are therefore not implemented.

A study based on text analysis

Mascha Kurpicz-Briki’s team used artificial intelligence in a way that automatically analyzes text and detects, on this basis, whether the language is burning. The goal is to study whether these free-text queries are effective and to create an automated metric to evaluate these queries. With success: The method accurately detects 93% of burnout cases. The scientist says:

“Automatic language processing is effective in detecting burnouts, although less time consuming, which is very promising.”

Data collection

As part of this work, the scientist and his team analyzed texts from the Reddit platform, an English-language community website that serves as a discussion forum organized by the theme. Although there is a subredit dedicated to burnout, the number of entries was too small to provide a sufficiently large dataset, so the team added text from a variety of thematic forums.

He created a dataset of 13,568 samples that described first-hand experiences, including 352 related to burnouts and 979 related to depression. He then used machine learning to create a method that evaluates whether a text is burnout.

Specifically, he classified the first collected text extract:

  • The text of the thread about burnout was manually categorized, with the exception of what Burnout mentioned.
  • Texts in other threads have been labeled as not related to mental health, not related to burnout.

Based on these examples, he has trained several models. Everyone has used different configurations to determine if a text (never seen by model) indicates a burnout. These models were then pooled as part of the diagnostic method, which proved to be very effective.

If these results are promising, then the cooperation of medical experts is especially necessary to verify the findings of this study on the real case of burnout in the next step and on the representative sample of the population. The data collected on Reddit is actually anonymous.

The authors point out that this work has been directed towards enhanced intelligence rather than artificial intelligence: instead of replacing clinical professionals, it seeks to embrace technology that empowers people in the decision-making process, providing information that is considered in human decision-making.

Source of the article: G. Merhaben, s. Nath, a. Putik, M. Kurpiz-Bricky: Burn-outNasemble: Augmented intelligence for identifying clues to burn-out in clinical psychology. Frontiers in Big Data (2022).

Leave a Comment