With depression, the quality of a person’s life is significantly reduced: the psychological state affects almost all its areas, from physical health to social interactions. However, even relatives and friends, not to mention the person himself, cannot always calculate depression. In this situation, the so-called neural networks can come to the rescue.
black and white world
According to psychologist Natalya Kiselnikova, words and phrases that people use in posts and messages can act as markers of depressive disorder.
“Analyzing the features of human speech –
- frequency of use of certain words,
- sentence length,
- lexical richness,
“Scientists can understand what mental state he is in,” she says.
When depressed, users are more likely to use adjectives with a negative connotation –
In addition, they are more focused on themselves than on other people and other things, and in their speech, first-person pronouns – “I”, “me” are found with great frequency, they rarely say: “we” or “they”. This is due not to selfishness, but to preoccupation with one’s own experiences.
Another sign of depression is the frequent use of the particle “not” in relation to oneself – for example, these are stories about their failures. And, finally, this is an extreme categoricalness in judgments: it is common for depressed individuals to use the terms
- “no way”.
“I get bullied all the time”, “no one here respects me”, “I never get lucky with money” – this is a typical lexicon in depression, when people see the world exclusively in black and white.
The first “psychotherapist” program called Eliza was created back in the 1960s by American programmer Joseph Weizenbaum. She simulated a dialogue with the client, asking questions based on his initial statements.
Suppose, if the client admitted that he had a headache, the algorithm was interested in: “Why do you say that you have a headache?” The interlocutor could answer: “My mother hates me.” Then the machine, in turn, asked: “And who else in your family hates you?”.
Although the questions were asked in simple patterns, many users of the program stated that they enjoyed talking with her and that she made them feel better.
Nowadays, there are self-help apps. For example, Woebot, Vivybot and Tess, which ask the user a series of standardized questions about their lifestyle, mood, habits, then analyze the answers and determine the presence of a disorder. If it is detected, the program gives recommendations, and in difficult cases it advises you to contact a specialist.
“Most of these bots do not contain what we call artificial intelligence at all,” explains Olga Kitaina, founder of one of the services for selecting psychologists. “They work on the basis of a “decision tree” that offers solutions based on scripts that are activated by your answers.” .
According to the expert, such applications can be used to support and popularize psychotherapy.
The most modern generation of neural networks uses arbitrary speech for analysis. These algorithms “observe” how people themselves use words, because the context of the use of a particular vocabulary can be individual.
The neural network “peeps”, for example, which words are combined with which most often and in what order in a person’s speech. The iCognito bot is able to detect the presence of a depressive or anxiety disorder with up to 87% probability, as well as recognize individual symptoms, such as
- feeling of loneliness
- negative attitude towards oneself
- social phobia,
- sleep or eating disorders,
- suicidal tendencies.
Not only words!
Experts believe that this kind of technology will not only help diagnose depression and other disorders in the early stages, but can also be used to educate people with autism: the presence of certain keywords in speech will help to recognize various social emotions. Also, the program will be able to determine if there is a deterioration in the condition, and whether the person intends to harm himself or even die.
Ideally, the algorithm should also take into account the peculiarities of speech depending on age and gender: after all, old and young people, men and women can express their state in words in different ways.
There are also neural networks that identify the psychological state from the images that users post or their network interactions.
Depressed people are more likely to post pictures in blues, grays and dark colors, and pictures with faces, but less often group pictures. However, not all users with depression behave in the same way, and even more deep learning is required so that algorithms can recognize symptoms with high accuracy.