Can Big Data and Machine Learning Help Us Crack the Dream Code?

While the use of big data in sleep science is widespread and well-documented, the more esoteric aspect of sleep – dreaming – has seen far less involvement from the big data industry. And that comes as no surprise. After all, this particular area has been far less studied scientifically and somewhat relegated to less-than-scientific approaches.

That being said, science has been investigating the phenomenon of dreams for a while and, more recently, we have even witnessed big data and machine learning being applied.

Dream Science

There have been scientific studies on dreams starting with the latter half of the 19th century, but they focused on psychoanalytics (most notably by Freud and Jung), which were lately mostly debunked.

They also relied on dream experiences as reported by the people taking part in studies. Still, there were some serious statistical studies done this way.

The true breakthroughs occurred in the 20th century with the invention of various diagnostic techniques such as:

  • Electroencephalography (EEG)
  • Electrocardiography (ECG)
  • Electronystagmography (ENG)
  • Electrooculography (EOG)
  • Electromyography (EMG)

These are combined to study people in the sleeping state in a procedure called polysomnography.

Another pivotal moment was the discovery of rapid eye movement sleep (REM sleep). During this sleep phase, the brain shows similar activity as in the waking state, and this is when most of the dreaming takes place.

Scientific dream studies involving polysomnography (and some other diagnostic methods) have gone a long way to explain the characteristics of dreams, such as their emphasis on the visual experience, the first-person nature, the lack of logic, and the strong emotions we often experience in dreams.

There have also been a number of theories proposing the reason why people (and some animals) dream, including:

  • No function – Dreams are the residue of waking neural action with no meaning or purpose.
  • Continuity – Our dream sequences follow what happens in waking life.
  • Psychological individualism – Dreaming reinforces a species’ typical behavior and contributes to a person’s individuality.
  • Emotional regulation – Dreams help us stay emotionally grounded and stable.
  • Memory consolidation – We dream to aid the memorization process.
  • Threat simulation – Dreams are there to help make us better prepared for threatening situations (hence so much running, falling, and conflicts in our nightmares).

Discovering the function of dreams is not the only reason why scientists are still trying to crack the dream code. Some believe that studying dreams can also help find treatments and medication for people struggling with mood, psychological and psychiatric disorders, PTSD, and other conditions.

Big Dream Data and Machine Learning

One of the biggest issues with historical studies of dreams had been the limited number of participants and dreams which could be used for any kind of research. More recently, there have been a couple of projects aimed at creating large databases of dreams.

The first of these, DreamBank, is a publicly available database of more than 24,000 dream reports, all collected over almost seven decades as part of scientific studies from around the world.

In August this year, a paper was published by a team of researchers who built an algorithm for the analysis of the entire DreamBank database, validated on hand-annotated dream reports.

With a success rate of up to 76%, the algorithm scored dreams for the ratio of positive-negative feelings, the level of aggression, and more. The team believes that this could be built upon, helping mental health professionals recognize potentially harmful developments by getting reports on their patients’ dreams.

Another project that hopes to create an even bigger dataset for dream analysis is the Shadow: Community of Dreamers app, founded by Hunter Lee Soik and featuring a team of data miners in fields like neurobiology to clinical psychology, from Harvard, MIT, Berkeley, and similar renowned institutions. The implications of having hundreds of thousands of dream reports from around the world are more than exciting for sleep and dream scientists around the world.

One of the most interesting applications of machine learning in studying dreams has to be a 2013 study in which a team trained linear support vector machines on fMRI data to try and find out if the visual cortical activity during REM sleep could predict what the participants dreamt. In simplest terms, the study showed that seeing things in our sleep matches the way in which we visually perceive objects in the waking state.

Closing Word

In a field where quantitative analysis is crucial to making new discoveries, big data and machine learning are bound to play a bigger role. Due to its (for the majority of people) more esoteric nature, dream science may be somewhat later to the party. But without a doubt, it will be further advanced by these approaches.

About the Author

Natasha Lane is a lady of a keyboard with a rich history of working in the IT and digital marketing fields. She is always happy to collaborate with awesome blogs and share her knowledge all around the web. Besides content creating, Natasha is nowadays quite passionate about helping small business to grow strong.

Sign up for the free insideAI News newsletter.

Speak Your Mind

*