Creating a personalized recommendation engine for managing burnout in employees

Singha S, Sharma L.K, Topdar A, Pradhan P.P, Srivastava R.K, Tatipally S

Burnout has become a prevalent cognitive phenomenon in today's workforce, hampering employee productivity. Recognizing its impact, the World Health Organization (WHO) classified burnout as an occupational hazard in the 11th revision of the International Classification of Diseases (ICD) in May 2019.  The complex, three-dimensional nature of burnout—emotional exhaustion, depersonalization, and reduced personal efficacy—requires targeted strategies to address each aspect. This project aims to develop a personalized recommendation system that suggests customized intervention journeys to employees based on their distinct burnout profiles.

The team analyzed burnout profiles, focusing on dimensions such as exhaustion, cynicism, and efficacy. Employees were classified into specific burnout profiles, including Low, Moderate, and High, based on their responses to the Maslach Burnout Inventory (MBI) survey questionnaire. Utilizing large language models (LLMs), the team generated generative text summaries for each user, which were then used to calculate cosine similarity with various intervention domain resources. The team employed TF-IDF, LDA, BERT Sentence Transformer Model, and OpenAI embeddings, with the best cosine similarity scores achieved using the OpenAI sentence transformer model.

For the recommender system, the team scraped data from YouTube based on extensive keyword mapping aligned with the MBI framework. A dynamic mechanism was implemented to calculate burnout scores based on user inputs, thereby classifying employees into their respective burnout profiles. The recommendation engine leveraged these scores and employee responses to suggest the most suitable intervention resources from the mapped categories.

Based on the project's findings, the team recommended that companies implement personalized recommendation systems to proactively address burnout. By continuously gathering employee input, the system can provide increasingly personalized intervention journeys, enhancing the effectiveness of the recommended interventions. Additionally, expanding the scope of intervention resources beyond YouTube videos to include other platforms can further tailor the recommendations to individual needs.

To enhance the recommendation system, the team recommends running A/B tests to experiment with different algorithms or strategies, such as TensorFlow. Collaborative filtering, which leverages the preferences of similar users for accurate and tailored recommendations, should also be implemented. Combining these techniques can continuously improve the performance and relevance of recommended videos, making the system more effective in addressing burnout.

Implementing a personalized recommendation system enables companies to address burnout proactively and promote employee well-being. Leveraging data analysis and intervention mapping can significantly improve employee engagement, job satisfaction, and overall performance. By focusing on customized interventions, organizations can mitigate the effects of burnout and create a healthier, more productive work environment.