Robotic Mental Well-being Coaches

My PhD thesis examines how robots could promote well-being. You can find my thesis here! It is open access with a CC BY-SA 4.0 license.

I conducted 4 user research studies (52 participants) to understand users’ needs and preferences, as well as professional well-being coaches’ professional advice and recommendations. The studies included interviews, groups discussions, and design workshops

I also conducted 5 real-world user studies (82 users) to evaluate users’ preferences and experiences with a robotic coach, and the success of the goal of the robotic coach. This included real-world deployments of the robotic coach at a workplace and a cafe.

The Task

This PhD research envisions robotic mental well-being coaches as tools to promote users’ well-being. With the rising prevalence of mental health disorders such as depression and anxiety, access to mental well-being services remains a significant challenge. Robotic mental well-being coaches are a promising approach for offering accessible tools to promote users’ well-being, potentially preventing mental health issues. This research set out to design and deploy a robotic well-being coach, which is designed according to users’ needs and stakeholders’ advice, and evaluate its effectiveness.

Overall, the PhD research aims to advance the human-centred and ethical design of robotic mental well-being coaches. It contributes insights and guidelines for researchers and developers, and an example of a design process which can be adapted to design future robotic coaches. The thesis contributes toward robotic coaches that can best serve their users, and have a positive impact by promoting users’ well-being in the real world.

The Solution

The effectiveness of robots for well-being hinges on being designed to meet users’ needs and concerns. As such, this PhD research used an iterative four-step Human-Centred Design (HCD) process applied to the design and deployment of such robots:

1. Understanding user needs and preferences for a robotic coach
2. Specifying design and ethical guidelines for the robot
3. Producing robots according to those guidelines
4. Evaluating users’ perceptions of the robots in real-world user studies

The Outcome

User Research

Prospective users and well-being professionals advocated for an accessible, usable, interactive and responsive robot that could conduct various well-being exercises. Prospective users called for robot capabilities such as acknowledging what they said, and robot features such as a form that matches the robot’s function. Professional coaches highlighted what kinds of well-being practices would be suitable for a robot to conduct (such as Positive Psychology and mindfulness meditation). Both groups raised and discussed ethical considerations such as privacy and safeguarding.

Design and Ethical Guidelines

Four design guidelines and three ethical guidelines for robotic mental well-being coaches were identified as a result of the analysis of user needs and well-being coach preferences.

They are:

• G1: Robot form should follow its function
• G2: Robot voice should emphasise variable prosody and slow pace
• G3: Acknowledgement and active listening should be implemented through backchannelling and appropriate turn-taking
• G4: Verbal adaptation should be limited to preserve well-being practice efficacy
• G5: Safeguarding and risk assessment are required prior to HRI
• G6: Data collection needs informed consent to preserve privacy
• G7: Researchers should educate users keeping in mind emotional considerations and user characteristics

Design Evaluation: User Study 1 at a Workplace

The first user study focuses on the examination of appropriate robotic coach form. We conducted a between-subjects user study (A/B test) to see if users preferred a toy-like or a child-like robot.

Participants preferred a toy-like robot over a child-like robot. Participants perceived the toy-like robot as having better behaviour (despite both robots behaving exactly the same) and a better personality, and they formed a stronger alliance with it. This may be due to a simpler robot form better matching the robot’s simple behaviour, and setting participants’ expectations of the robot’s capabilities at a more appropriate level.

However, participants had challenges taking turns with the robot, so we iterated on the robot’s turn taking skills in the second user study.

Design Evaluation: User Study 2 at a Workplace

Based on the findings from the first user study, the second user study distills insights for repair strategies for a robotic coach’s conversational errors. We conducted a four-phase study, including a consultation with a professional coach, a between-subjects user study examining empathic and non-empathic repair strategies, a design workshop with users, and a final consultation with the professional coach on the results of the study.

We found that it is not as simple as saying empathic or non-empathic repairs are better. Instead, in longitudinal interactions with users, robotic coaches should use different error repair strategies depending on context. For small errors, the robot may not need to repair or can use a brief apology. Some users had a personal preference for empathic repair strategies, and advocated for them especially in the case of severe errors. Robotic coaches should also use strategies such as technical explanations and instructions on robot use, which a human coach would not use.

Check out these publications related to the thesis:

• Axelsson, M., Bodala, I. P., & Gunes, H. (2021, August). Participatory design of a robotic mental well-being coach. In 2021 30th IEEE International Conference on Robot & Human Interactive Communication (RO-MAN) (pp. 1081-1088). IEEE. [Publisher]
• Churamani, N.*, Axelsson, M.*, Caldır, A., & Gunes, H. (2022, October). Continual learning for affective robotics: A proof of concept for wellbeing. In 2022 10th International Conference on Affective Computing and Intelligent Interaction Workshops and Demos (ACIIW) (pp. 1-8). IEEE. [Publisher]
• Axelsson, M., Spitale, M., & Gunes, H. (2023, March). Adaptive robotic mental well-being coaches. In Companion of the 2023 ACM/IEEE International Conference on Human-Robot Interaction (pp. 733-735). [Publisher]
• Axelsson, M., Spitale, M., & Gunes, H. (2023, March). Robotic coaches delivering group mindfulness practice at a public cafe. In Companion of the 2023 ACM/IEEE International Conference on Human-Robot Interaction (pp. 86-90). [Publisher]
• Spitale, M., Axelsson, M., & Gunes, H. (2023, March). Robotic mental well-being coaches for the workplace: An in-the-wild study on form. In Proceedings of the 2023 ACM/IEEE International Conference on Human-Robot Interaction (pp. 301-310). [Publisher]
• Spitale, M., Axelsson, M., Kara, N., & Gunes, H. (2023, August). Longitudinal Evolution of Coachees’ Behavioural Responses to Interaction Ruptures in Robotic Positive Psychology Coaching. In 2023 32nd IEEE International Conference on Robot and Human Interactive Communication (RO-MAN) (pp. 315-322). IEEE. [Publisher]
• Axelsson, M., Spitale, M., & Gunes, H. (2024). Robots as mental well-being coaches: Design and ethical recommendations. ACM Transactions on Human-Robot Interaction, 13(2), 1-55. [Publisher]
• Axelsson, M., Spitale, M., & Gunes, H. (2024, March). “Oh, Sorry, I Think I Interrupted You”: Designing Repair Strategies for Robotic Longitudinal Well-being Coaching. In Proceedings of the 2024 ACM/IEEE International Conference on Human-Robot Interaction (pp. 13-22). [Publisher]
• Spitale, M., Axelsson, M., & Gunes, H. (2023). Vita: A multi-modal llm-based system for longitudinal, autonomous, and adaptive robotic mental well-being coaching. ACM Transactions on Human-Robot Interaction. [Publisher]
• Spitale, M., Axelsson, M., & Gunes, H. (2024). Appropriateness of LLM-equipped Robotic Well-being Coach Language in the Workplace: A Qualitative Evaluation. 2024 33rd IEEE International Conference on Robot and Human Interactive Communication (ROMAN), 2024, pp. 1571-1578. [Publisher]
• Axelsson, M., Churamani, N., Çaldir, A., & Gunes, H. (2025). Participant Perceptions of a Robotic Coach Conducting Positive Psychology Exercises: A Qualitative Analysis. ACM Transactions on Human-Robot Interaction. [Publisher]