Privacy-Preserving AI for Mental Health Detection in Care Robotics

About the Project

Project Description:

The College of Business, Technology and Engineering draws on talents, expertise and facilities across Sheffield Hallam University. The vision is to be the leading provider of applied research excellence delivering business, materials, computing, science and engineering innovations meeting the development needs of industry.

This project is part of a Graduate Teaching Assistantship scheme, in which the successful applicant will undertake certain teaching duties associated with the student experience, in addition to working towards a PhD qualification. They will contribute up to 180 hours of support for research or teaching related activity per academic year. This activity forms part of the scholarship and there is no additional payment.

PhD Research Topic:

The rapid evolution of care robotics and artificial intelligence (AI) is transforming healthcare delivery, particularly in areas requiring long-term monitoring, personalised support, and mental-health assistance. Socially assistive robots are already demonstrating measurable benefits in elderly care, rehabilitation, and emotional well-being, while reducing pressure on healthcare professionals. At the same time, Large Language Models (LLMs) have revolutionised conversational AI, enabling natural, adaptive dialogue and offering new possibilities for mental-health support through accessible and anonymous interactions.

Despite these advances, significant challenges remain. Current AI-driven systems often lack trustworthiness, robust privacy safeguards, and clinically validated response mechanisms, especially when handling sensitive mental-health data. Concerns such as misinformation, hallucination, emotional dependency, and data leakage highlight the urgent need for safer and more ethical AI solutions. Addressing these issues is critical for enabling the responsible deployment of AI in healthcare.

This PhD project aims to tackle these challenges by developing a privacy-preserving, AI-based LLM system capable of detecting anxiety and depression while ensuring safe, trustworthy, and clinically aligned interactions. The importance of this research is underscored by the growing mental-health crisis in the UK: approximately 1 in 6 adults experience mental-health problems, with even higher prevalence among young people and individuals with disabilities. At the same time, NHS services face increasing strain, with millions of patients experiencing long waiting times for care. This project seeks to complement existing services by enabling accessible, real-time mental-health support through AI-enabled care systems.

The core research question is: How can we design a safe, trustworthy AI system that preserves user privacy while accurately assessing mental-health conditions, achieving standards comparable to clinically validated practices?

To address this, the project will pursue several key objectives. These include conducting a comprehensive review of AI applications in mental healthcare, developing and fine-tuning a lightweight LLM for anxiety and depression detection, and designing a role-based privacy framework that controls information access depending on the user (e.g., patient, clinician, caregiver). The system will also incorporate clinically guided responses aligned with recognised therapeutic approaches and NHS pathways.

Methodologically, the research adopts a multi-layered approach. It begins with ethically sourced and carefully processed mental-health datasets, ensuring compliance with GDPR and ethical standards. A lightweight LLM (e.g., GPT4All-LoRA) will then be fine-tuned to detect mental-health indicators efficiently, even on resource-constrained devices. A novel privacy architecture will be implemented to enforce role-based access, prevent sensitive data leakage, and ensure secure interactions. Finally, the system will be deployed on both advanced robotic platforms (such as Care-O-Bot) and embedded systems (e.g., Nvidia Jetson), enabling evaluation across real-world scenarios.

This PhD offers an exciting opportunity to work at the intersection of AI, robotics, privacy, and healthcare, developing technologies that are not only innovative but also ethical, scalable, and socially impactful.

Eligibility

Applicants should hold at least a 1st or 2:1 Honours degree in AI and/or/with Cybersecurity or a related discipline.

We strongly encourage applications from individuals from groups underrepresented in postgraduate research, including but not limited to women, LGBTQ+, and minoritised ethnic groups.

Information for international applicants

English language requirements of IELTS 7 with a minimum score of 6.5 in all test areas (or equivalent) are mandatory if English is not your first language. Qualifications should have been taken within the last two years. Further information can be found here.

How to apply

To apply for this GTA scholarship, please use our online application form.

You must ensure that you upload:

1. A personal statement (up to 2 pages maximum) detailing your interest in the project and how your experience in academia, industry, research or social activities makes you the best candidate (Please upload this in place of a proposal). We’re looking for evidence of:
   • motivation and curiosity for postgraduate research
   • analytical and technical expertise related to the research proposal
   • ability to communicate clearly
   • planning and organisational skills
   • ability to work independently and collaborate with others
   • commitment to integrity and responsible research
   • resilience to setbacks and challenges
   • where you might contribute to teaching
2. A two page (maximum) CV
3. Two letters of reference, or details of two referees, at least one from an academic and both dated within the last 2 years
4. Copy of your highest degree certificate and transcript. If your degree is not yet awarded then submit a copy of your latest transcript.
5. Non-UK applicants must submit IELTS results (or equivalent) taken in the last two years and a copy of their passport

If you are applying for multiple GTA projects, please clearly list them all in your application. You will need to submit a tailored personal statement for each project.

Application deadline: 07 May 2026

Start date: October 2026

Interviews: TBC

Information about our research degrees can be found here.

Funding Notes

The GTA scholarship is for 3.5 years of full-time study and provides tuition fees at both the UK (home) and international level plus a maintenance bursary in line with guidance from UK Research and Innovation and the Living Wage Foundation (for illustrative purposes, the Sheffield Hallam University bursary for 25/26 is £22152). GTA scholarships are open to both UK (home) and international applicants.