Domestic Data Streamers has utilized the technique to preserve memories of individuals in various migrant communities, such as Korean, Bolivian, and Argentine families residing in São Paolo, Brazil. They have also collaborated with a care home in Barcelona to explore how memory-based reconstructions can benefit older individuals. Working with researchers in Barcelona on a small pilot involving 12 subjects, the team applied the approach to reminiscence therapy, a treatment for dementia that aims to stimulate cognitive abilities through visual stimuli of past images. Reminiscence therapy, developed in the 1960s, has varying opinions on its effectiveness and execution.
The pilot enabled the team to refine the process and ensure that participants could provide informed consent, according to Garcia. They are now planning a larger clinical study in the summer with colleagues at the University of Toronto to compare the use of generative image models with other therapeutic methods.
Through the pilot, it was discovered that older individuals had a stronger emotional connection with printed images compared to viewing them on a screen. Garcia notes, “But when they could see it physically, the memory got much more important.”
Blurry is best
The researchers have found that older versions of generative image models are more effective than newer ones. Initially using DALL-E 2 and Stable Diffusion, they later switched to Midjourney but found that the results did not resonate well with individuals.
Garcia explains, “If you make something super-realistic, people focus on details that were not there…Memories are a bit like dreams. They do not behave like photographs, with forensic details.”
The team has reverted to using the older models, with Garcia stating, “For us, the glitches are a feature…It’s kind of a quantum state in the images that works really well with memories.”
Independent filmmaker Sam Lawton, not affiliated with the studio, is intrigued by the project and the upcoming clinical study on the cognitive effects of generated imagery. Lawton has used generative image models to recreate his own memories in his film, Expanded Childhood, blending real childhood scenes with surreal ones.
Lawton shares, “The effect exposure to this kind of generated imagery has on a person’s brain was what spurred me to make the film…I pivoted to the kind of storytelling that’s most natural to me.”