Rememo
In this research-through-design study, we explored reminiscence therapy, which uses memory triggers (e.g., photos, objects, smells) to invite storytelling with people living with dementia. Yet in institutional settings in Singapore, such triggers are often unavailable, hard to source, or not relevant to the individual. Time-strapped therapy staff—some of whom are non-local—also face language and cultural barriers when preparing and facilitating sessions. We therefore asked how generative AI might function as one material among many within care work to support facilitators.
Rememo pairs illustrated prompt cards with a mobile web app for generating culturally grounded images for discussion. Staff and residents browse cards with translations and pronunciation cues; staff then scan selected cards to generate an image situated in the old Singapore context, review the output, and print it as a physical photo for use in conversation. By embedding AI within a tangible, facilitator-led workflow, Rememo extends staff capacity while keeping care human-led.
Rememo also frames these prompts as synthetic memory, not to implant false memories, but to craft “handles” that make remembering and storytelling easier as we shift AI in care towards an AI-in-the-loop model accountable to therapeutic practice.
Celeste Seah
Yoke Chuan Lee
Jung-Joo Lee
Ching-Chiuan Yen
Clement Zheng
We began with field visits to dementia care facilities in Singapore, using ethnographic observations and informal conversations with therapists to understand how reminiscence therapy (RT) is prepared and facilitated. Through a two-year research-through-design process, we iteratively developed the Rememo system by balancing contextual inquiry, system design, and AI integration. Working closely with therapy staff throughout the project, we eventually user-tested the system in two nursing homes. This allowed us to observe how Rememo fits into real-world therapy workflows.
Stock materials currently used in RT lack cultural relevance or personal resonance. While they may spark interaction in some, they fail to truly evoke memories or emotional engagement for other clients. This reduces the therapeutic potential of the session and places greater burden on therapists to fill in the gaps.
Understanding how to trigger meaningful memories often depends on therapists’ accumulated tacit knowledge—unwritten, intuitive insight built over time through repeated interaction. This knowledge is difficult to externalize or transfer, making it challenging for new staff or caregivers to support reminiscence therapy at the same level of depth.
Due to a lack of manpower in the care sector, many of the therapy aides hail from neighbouring countries and do not share the same cultural background as the clients. Moreover, some clients also only speak in the local languages and dialects that foreign staff struggle to understand.
In contrast to systems that aim to replace human facilitators, Rememo is a therapist-oriented tool that follows an AI-in-the-loop model. We found that by balancing autonomy between therapists and clients, the system empowers both to share control over the conversation. Coupled with generated visual aids relevant to session, clients are more engaged and encouraged to share richer, personally significant stories.
By providing contextually relevant guiding questions with each generated image, the system helps novice facilitators overcome the initial hurdle of leading a session. This is designed to ease them into the conversational flow of therapy while they learn to interpret subtle resident cues. Over time, this support allows staff to gain confidence and hone their own facilitation expertise.
The physical design of the system includes large prompt cards featuring illustrations and prompts in multiple languages, ensuring the tool is accessible to both therapists and residents. By incorporating these multilingual cues, the cards help minimize language barriers, facilitating clearer communication and more inclusive participation during therapy sessions.