Thesis
Establishing Design Principles for Human-AI Interaction
Establishing Design Principles for Human-AI Interaction
"Throughout his studies, Shane demonstrated an exceptionally high level of professionalism, enthusiasm and energy for his studies. He engaged at an excellent level, going above and beyond the demands of each project brief. Over the year, I was consistently impressed by his intellectual curiosity and dedication to his work."
Emma Creighton, Interaction Design Director, NCAD
This project was the focus of my Master’s thesis and was developed in partnership with Accenture, specifically their Human Science Studio. The brief asked: “How might we design inclusive AI companions for the elderly and vulnerable?” with a parallel focus on: “What ethical considerations should be taken into account when designing AI for care settings?” Given the growing role of AI in healthcare and the urgent need for ethical, human-centred design in this space, I was drawn to the relevance and complexity of this challenge. I was particularly excited to explore it with the support and mentorship of Accenture, whose expertise and resources helped ground the project in both practical and speculative futures. My approach centred on designing provotypes—provocative prototypes intended to spark conversation and reflection with experts in dementia care. These engagements were used to surface tensions, challenge assumptions, and collaboratively define a set of inclusive design principles that could guide future work within Accenture’s innovation hub, The Dock.
Memoira is a speculative AI companion designed to support individuals with dementia through reminiscence therapy—a conversational approach that encourages users to recall and reflect on meaningful life moments. Using conversational AI, Memoira gently guides users through memory-sharing prompts. As a memory takes shape, generative AI visualises the scene and prints a tangible keepsake, reinforcing memory recall and deepening emotional engagement. Designed to foster companionship and stimulate cognitive activity, Memoira doesn't replace human care—it enhances it. By turning personal stories into shared experiences, it creates space for meaningful, memory-driven connection.
To ground the project in real-world needs, I began with both primary and secondary research—exploring the limitations of AI in care settings and the unique challenges faced by people living with dementia. These insights informed the development of a provotype (a provocative prototype) designed to surface ethical tensions and spark dialogue. Using this concept as a conversational tool, I facilitated discussions with dementia care experts to co-create a set of inclusive design principles that could guide the development of AI companions in sensitive, real-world contexts.
To ground the project in real-world context, I conducted a mix of academic research and expert interviews across AI, healthcare, and elder care. This revealed key tensions: a growing demand for AI due to critical staff shortages, alongside deep concerns about its emotional impact on people with cognitive impairments. Experts stressed the importance of designing AI that empowers—supporting independence while maintaining empathy, transparency, and trust.
My initial explorations resulted in two distinct provotypes: MyNurse+, an AI-powered assistant designed to support daily living in nursing homes, and M-STAR Bot, a speculative companion aimed at stimulating memory and capturing personal stories from individuals with dementia. Leveraging SoulMachines, I quickly built and tested a functional avatar for MyNurse+, gathering early feedback from stakeholders. However, based on initial insights and the limited time available, I recognised that this concept would require a longer-term exploration to do it justice. To make the most impact within the project scope, I chose to focus my efforts on developing M-STAR Bot—which ultimately evolved into Memoira, the core outcome of this project.
To explore how Memoria could take shape, I rapidly prototyped multiple iterations—starting with its physical form. I grounded the design in familiar, nostalgic technology that would resonate with older users. Early versions took inspiration from vintage radios, but through testing and iteration, the form naturally evolved into something more symbolic: an old Polaroid camera—an object closely tied to memory and storytelling. To bring Memoria to life, I integrated Arduino components to create a more interactive, tangible experience. On the software side, I developed a custom GPT using ChatGPT, paired with DALL·E. This conversational AI would ask users guided questions about a childhood memory, generate a visual interpretation of the story, and print it using a Bluetooth-connected Polaroid printer
I began by testing the prototype with a broad age range—including friends, family, and as many older users as I could reach. While this helped validate the basic functionality and usability of the experience, the feedback lacked the depth needed to shape meaningful design principles. It became clear that to truly understand the ethical and emotional impact of Memoria, I needed to engage directly with experts in dementia care and elder support.
To gather deeper, more informed insights, I visited two memory clinics in Dublin—St. James’s Hospital and the Mater Hospital. These sessions enabled rich conversations with medical professionals who work closely with dementia patients. From these discussions, three key themes emerged: Ethical Considerations: Clinicians emphasised the importance of safety in AI systems, especially those monitoring patients. The ability to detect signs of distress and alert caregivers was seen as critical. Design Considerations: The slower cognitive and response times of users with dementia called for an interface that is patient and gentle in tone—prioritising calm, easy-to-follow interactions. Severity of Dementia: Doctors expressed serious ethical concerns around late-stage dementia patients potentially misinterpreting the AI as a real human, underscoring the need for clarity and transparency in all interactions. Using these insights, I developed a set of draft design principles to guide ethical and inclusive AI interactions. To validate and refine them, I returned to Accenture—my project partner—and facilitated a final workshop with the Human Science Studio. Their input helped sharpen the principles, ensuring they were not only grounded in field research but also aligned with industry standards for responsible innovation.
