View all projects

Agent for emotion regulation and behavioral activation

Rasa
Node.js
React

Nowadays, patients affected by disorders such as depression or anxiety are generally followed by professional therapists. After these sessions or a stay in a clinic, some patients may relapse. The development of an IT solution to monitor, assist and advise these people seems essential to support them in their efforts.

Therapy session

Conversational agents are particularly well suited to this type of task. They are gaining in importance in modern IT systems and are evolving very rapidly.

Goals

The main objective of this project is to develop a conversational agent to help sick people manage their emotions. The system's interface must be simple and intuitive to use, but above all accessible to anyone, in accordance with current standards in terms of accessibility.

Work done

Following the analysis of the literature on emotion regulation and conversational agents, a web platform was developed. It is inspired by acceptance and commitment therapy, and more specifically by behavioral activation. The two main functionalities are the definition of personal goals and the establishment of a daily plan.

Three screenshots showcasing the app interface

The user can interact with the conversational agent, named Lucky, directly via the platform's accessible and responsive web interface. The actions performed by the user through the interface have repercussions on the course of the conversation with Lucky. The latter will constantly suggest to the user what he can do: set new goals, plan his day, or express his feelings. These suggestions are made in an intelligent way and in synchronization with the database.

Validation

In addition to automated tests, ten volunteers took part in an in-depth evaluation of the platform, during which they had to complete four imposed tasks, and then answer a questionnaire about the interaction with the chatbot and their personal feelings.

With an average score of 81, the usability of the conversational agent is already very good, even if several points of improvement were retained. Overall, the participants are open to the idea of using this application on a daily basis, whether it is in addition to another application or not. Even if these people do not suffer from psychological problems, the results obtained give an initial positive overview of the users' feelings towards the platform.

Footnotes

I worked on this project for my Master thesis at the HES-SO in Lausanne. This project was given a grade of 6 / 6.