AI is helping scientists interpret data from dream reports that have the potential to produce indicators for mental health and overall well-being. (Credit: Getty Images) 

By AI Trends Staff  

Scientists are developing AI tools to help analyze dreams, in the hopes of better understanding where dreams come from and helping people address real-life problems, especially around mental health.   

Scientists in the UK and Italy have created an AI tool to analyze dream reports, which are text reports written by the dreamer when they wake up. The analysis of dream reports previously demanded a time-consuming manual annotation of text, which is why dream reports have recently been mined with algorithms focused on identifying emotions, according to a recent account in the Royal Society Open Science journal.  

The goal is to mine important aspects of dream reports, such as characters and interactions, in a principled way grounded in academic literature. The team designed a tool that automatically scores dream reports based on a widely-used dream analysis scale. Results so far have been encouraging. 

Dr. Luca Maria Aiello, senior research scientist, Nokia Bell Labs

Based on a scoring of 24,000 dream reports, “We found supporting evidence that dreams are a continuation of what happens in everyday life,” said the authors, which included Dr. Luca Maria Aiello, a senior research scientist at Nokia Bell Labs. “Our results suggest that it is possible to quantify important aspects of dreams, making it possible to build technologies that bridge the current gap between real life and dreaming.” 

Dream Reports Have Potential to Produce Mental Health Indicators 

The dream reports were taken from DreamBank.net, the largest online public repository of written dream reports, with over 38,000 dream descriptions gathered from a variety of verified sources and research studies. 

The researchers built a tool that applied constituent-based analysis, a technique used to break down natural language text into its constituent parts that can then be later analyzed independently. The result of this procedure is a parse tree, a dendrogram whose root is the initial sentence. The tool incorporated the Stanford Parser from a Python toolkit, a widely-used, state-of-the-art parser. The tool outputs the parse tree and annotates nodes and leaves.  

The practical implications of the research, according to the report, are to make it possible to build technology that automatically classifies dream reports. “This could result in automatic diagnostic or prognostic indicators for mental health in general,” the authors stated. 

Dream Reports Seen Able to Shed Light on Effect of Pandemic 

Psychologists could also use the information to understand in real-time how events such as wars, natural disasters and even a coronavirus pandemic are impacting people’s mental health, suggests an account in inews.uk.  

“We hope more and more people will have an incentive to share their dreams,” said Dr. Aiello. “We would be able to run this dream analysis at large scale and understand if global events such as pandemics, wars, financial crises and even global warming have an effect on our psychological well-being as reflected in the dreams that we dream.” 

Researchers did research dreams after the 9/11 terror attack in the US. “Psychologists saw there was a correspondence between bad things happening in people’s dreams in the US,” Dr. Aiello stated. “Imagine if this could be scaled up with many more global events.” 

He added,  “We are not trying to replace the experts, but we are trying to automate part of this process to make it easier for them. At large scale, you might imagine that events like the Covid-19 pandemic are most likely affecting how people feel and affecting their dreams.”  

Finally, “If most people were to share those dreams we could monitor over the different weeks of the pandemic, how people feel about it. In principle, if enough people shared their dreams, this could be done in real time.” 

Berkeley Scientists Working on Creating Video From Dreams 

Other ways to try to register dreams using AI have been tried. In 2011, researchers at the Gallant Lab at the University of California, Berkeley, had participants watch movie trailers while monitoring their brain activity, according to an account in Discover. The researchers constructed low-resolution videos of what the participants were watching using only their brain activity. They improved on the process and published a follow-up study in 2016. The reconstructions were rough patterns and not high-definition, but the experiment piqued the interest of scientists interested in trying to record dreams.  

Martin Dresler, Principal Investigator, Donders Sleep and Memory Lab in the Netherlands

“We don’t know for sure, but some day, I could imagine that it’s possible [to build a dream video ]. The information that represents the dream is present somewhere in the brain, so in principle, there’s no reason why it shouldn’t be possible,” stated Martin Dresler, Principal Investigator at the Donders Sleep and Memory Lab in the Netherlands.  

Yukiyasu Kamitani, a neuroscientist at Kyoto University who is conducting dream research falling at the intersection of neuroscience and computer science, expressed a similar sentiment. “I think at least some visual aspects of dreaming can be captured in the form of a movie at low spatial and temporal resolution. It should be noted that a movie might be just a coarse approximation for a dream,” stated Kamitani.  

Scientists at the University of Wisconsin-Madison are studying which region of the brain is activated when dreaming is taking place. In April 2017, the scientists identified a “posterior cortical hot zone” where when dreams were occurring, high frequency activity was measured in the zone, and low frequency activity decreased. The results suggest an ‘on-off’ switch for dreaming can be observed.  

The researchers also found that parts of the brain involved in perception during waking hours, behave the same way during sleep. “When we wake someone up, and they report hearing something, or there was speech for example, we find activation in a very specific part of the cortex: the Wernicke’s area, which is known for processing speech,” stated Benjamin Baird, a lead scientist on the study. “When people reported seeing a face, we found activation in a very specific part of the brain, known as the fusiform face area. So it appears that specific types of perceptions during dreams activate the same brain regions as perception during wakefulness.” 

Read the source articles in the Royal Society Open Science journal, in inews.uk and in Discover.