A new experimental AI machine has been developed by a team of three researchers in order to predict more accurately if a person will have psychotic breaks.
This is a huge step forward for medical science.
The team, consisting of Neguine Rezaii of Harvard Medical School and Emory School of Medicine, and Elaine Walker and Philipp Wolff from Emory's Department of Psychology, developed the machine that can listen for early whispers and signs of a psychotic break - not something the human ear can do.
What does this discovery mean?
A machine learning method was built to look for the specific indicators typically associated with psychosis, and specifically schizophrenia.
It took the team two years of studying volunteers' data and reactions. Most volunteers ended up having psychotic breaks.
This is an incredible feat.
The team was able to demonstrate that by using their experimental machine they could predict psychotic breaks much more accurately than humans could. More than that, they also discovered that their invention could predict if a person is experiencing early signs of auditory hallucinations.
The latter could have worldwide implications for the field of psychology as it is a new indicator for impending psychotic breaks through auditory hallucinations. If discovered early enough, especially in the case of schizophrenia, this could mean better, more specific medical treatment.
It has to be pointed out, however, that there is currently no cure for psychotic breaks. However, by discovering potential psychotic breaks at an earlier stage better aligned care and treatment are possible for the patient. Making their lives, as well as those caring for them, a little easier.
The team's discovery in AI could make a huge difference in the care for those suffering from future psychotic breaks.
How does the machine work?
As per the team, "Our findings indicate that during the prodromal phase of psychosis, the emergence of psychosis was predicted by speech with low levels of semantic density and an increased tendency to talk about voices and sounds. When combined, these two indicators of psychosis enabled the prediction of future psychosis with a high level of accuracy."
Those who speak with low semantic density communicate differently, they have to be prompted when speaking, and don't offer much substance in their answers.
The difference between low, or Alogia, and regular semantic density can be seen through this example:
The tricky part of the research was discerning whether or not people were hearing voices or sounds in their heads that don't in fact exist.
The team created a technique called 'vector packing', which means finding out how much information is packed into one sentence.
In the end, the team discovered that those who wound up having psychotic breaks were those who mostly used words associated with sounds and noises in their descriptions.
When combining their two methods, of semantic density and vector packing, the team had a 93% accuracy rate of knowing who would have psychotic breaks.
This positive discovery may have extremely positive implications for the field of mental health, providing more targeted treatment.
It shows just how useful machine learning is, and what incredible impacts it can have on AI research, medical science and human history.