Sensing the Past – Memory, AI and Trauma
Introduction
This work-in-development — ‘Sensing the Past’ is an extension from my current art project exploring memory, trauma and artificial intelligence (AI). Galvanising Change, which explores climate justice and trauma, uses AI seeking to disseminate its ideas whilst not re-traumatising its audience.
Initial Research
During the research period for ‘Sensing the Past’ I discovered that AI and machine learning have been used for narrative exposure therapy in an attempt to avoid re-triggering the patient through asking them to recall and remember traumatic memories.
Ref. Olsen, A., 2018. New Trauma Therapy using storytelling, music, artificial intelligence (AI) and virtual reality (VR). [online] Medium. Available at: <https://medium.datadriveninvestor.com/new-trauma-therapy-using-storytelling-music-artificial-intelligence-ai-and-virtual-reality-vr-3970bf361937> [Accessed 1 September 2022].
The research process for ‘Sensing the Past’, and this knowledge exchange project, both seemed to be the perfect means for extending the concept for ‘Galvanising Change’ whilst exploring memory and trauma.
After my initial research and an investigation of various AI off-the-shelf tools I chose two – Melobytes and Night Cafe Studio. Melobytes takes an uploaded photograph/image, and uses the imagery from it to produce a soundscape, or a music track. The ‘Night Cafe Studio’ tool produces an AI processed image from combining a text prompt and a base image.
To investigate this AI visualisation process I decided to use a set of photographs I originally used as inspiration to recall the memories of each traumatic incident and to write the 12 audio stories for ‘Galvanising Change’. During the writing process for each of the six stories I wrote an intense and not intense version, drawing from my memories of the incident. For example Story 1B recounts walking alone in the Cairngorms aged 14 and seeing a golden eagle, the emotions in the memory are of fear of being alone, that I might trip and injure myself and not be found or get down the mountain before nightfall, and that the eagle might attack me. Story 1A is a less intense version of the same story but written in the third person, instead of the first person, with less graphic details of the incident. I also used grammar, such as adjectives to affect the intensity.
Taking the ideas from the ‘New Drama Therapy’ article I experimented with using AI with a paragraph of text from each audio story to automatically evolve the photo. I used the intense version of each, e.g. for Story 1B I extracted a paragraph that contained the emotions of the traumatic incident. This was repeated for each of the six climate justice audio stories.
The resultant photographs appear to almost bring to light the memories and ideas behind the photographs, it’s as if the AI could understand the trauma and emotions behind these paragraphs of text and use them to create a visualisation of the memory.
Original Memory Images
AI Memory Images
In addition to this experiment in visualising emotions from photographs and audio stories, and this perhaps being an indication of their emotional impact, this was a really useful process for me to understand the emotional impact of the words I have written, particularly as audio and voice act different on the brain, and how they in turn can trigger memories, perhaps trauma in the audience.
Going forward I would like to understand the mechanisms by which the machine learning algorithm has manipulated the pixels in the source photograph. I have a basic understanding of sentiment analysis, which classifies words as good, bad or neutral in order to model the sentiment of a paragraph of writing. It would be good to have the mechanisms by which to obtain a more nuanced analysis of the photographs and be able to apply that algorithm to a photograph to produce a memory visualisation.
Using Melobytes and the same series of photographs the AI produced soundscapes, again, I feel, making visceral the traumatic aspects of each incident.
The combination of sound and image brings to the fore the impact of both seeing and hearing what could be described as disturbing images / sounds, again anticipating what the impact on the audience would be.
“In traumatized people, memories can be a jumbled mess of scary pictures scenes, and images. Second is the emotional aspect which has to do with the way you feel when you think about your experience (fear, anguish, pain, etc..). Traumatized people cannot completely detach themselves from their traumatic memories so may still have fresh emotions when they remember them. The third is the somatic aspect which is the way your body reacts to your feelings like trembling, sweating, agitation, and so on. When you experience a traumatic event, you keep on reliving the moment because the trauma affects parts of your brain like the forebrain (controls cognitive), limbic system (controls emotional), and hindbrain (controls somatic) which help archive your memories. What this means is that either all or part of the memory is not properly archived into the long-term memory storage. Trauma disrupts the cognitive, somatic, and emotional aspects of your memory. This means that anytime you remember such an event, it is as if it is literally happening to you.” (Olsen, 2018)
All work featured in the Creative Lives Gallery 2022 belongs to the artists and can only be reproduced with their permission.
Creative Lives 