Andrea Buccoliero

PhD programme in Human Sciences - 38° ciclo (October 1, 2022 - September 30, 2025)

Doctorate research program

VOICE BIOMARKERS OF BURNOUT SYNDROME
An Artificial Intelligence (AI) agent prototype to infer Burnout risk via human speech

One of the fastest-growing domains, understood in terms of epidemiological weight and social and organizational impact, is that relating to Mental Health as an integral part of health and well-being as defined by the World Health Organization (WHO) Constitution. Among the various priorities related to Mental Health, facing Burnout syndrome is one the most challenging. In May 2019, burnout has been recognized as a syndrome and, as such, has been listed in the 11th revision of the International Classification of Disease (ICD), the global reference text for all diseases and health conditions. The World Health Organization defines burnout as an "occupational phenomenon" resulting from poorly managed chronic stress (WHO, 2019).

The project aims to identify biomarkers of Burnout in the voice of individuals, and to develop an AI Agent able to perform the assessment. It spans several application domains, all intertwined by using speech as a proxy for accessing human emotional state. The key idea is to adopt modern Speech Emotion Recognition (SER) (Schuller, 2018) algorithms to evaluate burnout presence and burnout risks among individuals.

During the initial part of the PhD programm, the project took the strand of literature research to have a clear comprehension of the state of art on the topics: Speech Emotion Recognition (SER) & Burnout syndrome Assessment.

On the SER side, Khalil, Jones, Babar et al propose a review of Speech Emotion Recognition Using Deep Learning Techniques. The primary goal is to comprehend human emotional state. Efforts are required to enhance the accuracy of emotion recognition by machines. On the Burnout side, this syndrome is currently measured through several Patient-Reported Outcome Measures (PROMs) and some of them have become widely used in occupational health research and practice. To be validly and reliably used in medical research and practice, PROMs should exhibit robust psychometric properties. Among the different PROMs, Cognitive behavioral intervention (CBI) and, to a lesser extent, Oldenburg Burnout Inventory (OLBI) meet this prerequisite. (Y. Shoman, 2021).

According to the introduced SoA, this project aims to address the following research questions:

1. What is the current view of Burnout? Do authors include in their discussion the role of voice?
2. What is the empirical evidence on Speech Emotion Recognition accuracy? How can we measure it?
3. What are the implications on devising assessment interventions for individual wellbeing?
4. Are there any Explainable Burnout Biomarkers hidden in the voice?
5. Can we suggest screening strategies to monitor the phenomenon in a sustainable and privacy-compliant way?

To answer the above-mentioned questions, this project is divided into three main studies.

The first study consists in an overview of the existing methods, tools, and models to assess Burnout risk via SER. The second study is an empirical work devised to design & develop applications and to collect data. The third is to design the AI agent and to test it.

Doctorate reference figure

Riccardo Sartori

PhD Tutors

Andrea Ceschi