RetroMind and the Image of Memories: A Preliminary Study of a Support Tool for Reminiscence Therapy

Reminiscence therapy, a widely used method focused on recalling past memories, has shown good results in improving the well-being of dementia patients, particularly with patients with Alzheimer’s disorder. However, despite promising results and wide use, its effectiveness remains a matter of debate. This paper presents RetroMind, a new framework that combines Large Language Models (LLMs) and social robots to improve the effectiveness of reminiscence therapy through the creation of visual material based on memories. RetroMind leverages the natural linguistic abilities of LLMs and the physical presence of social robots with the goal of engaging patients in personalized reminiscence sessions. The main goal of RetroMind is to provide an image consistent with the patient’s narrative that serves as a support for reenactment and consolidation of memories, improving the effectiveness of therapy. The framework incorporates two questionnaires, the Autobiographical Memory Interview (AMI) to assess autobiographical memory and the Cornell Scale for Depression in Dementia (CSDD) to assess patients’ levels of depression, enabling targeted and individualized interventions.