Why Do We Measure Blood Pressure but Not "Changes of the Heart"?—Early Detection of Mental Health in the AI Era

When you go for a health check-up, you measure your blood pressure, undergo a blood test, and check your blood sugar and cholesterol levels.

Few people question this routine.

If your blood pressure is high, you reconsider your lifestyle. If there's an abnormality in your blood sugar levels, you undergo further testing. You detect signs before they develop into serious illnesses and seek treatment if necessary.

This approach of "detecting symptoms before they become severe" is a common practice in modern medicine.

However, when it comes to mental health, the situation changes significantly, even though it's also a health issue.

You experience sleepless nights. You don't want to go to school or work. You avoid social interactions. You lose interest in things you once enjoyed. You become extremely pessimistic about the future.

Even when these changes occur, it's not uncommon for those around you to notice only after your grades have plummeted, you've started taking long absences, or your work performance has significantly declined.

In other words, while we emphasize "prevention" for physical health, the structure of "dealing with issues after they surface" still remains for mental health.

To bridge this gap, discussions have begun on the combination of regular mental health screenings and AI.

Are School Safety Measures Focused Only on the "Last Few Minutes"?

This issue is particularly significant in schools.

When a major incident occurs at a school, the first discussions naturally revolve around physical safety measures.

Strengthening inspections at school gates. Increasing security personnel. Creating systems to prevent dangerous items from being brought into the school.

These measures are necessary.

However, there are problems that cannot be addressed by these measures alone.

The method of detecting dangerous items at the school gate is, in other words, a system that attempts to stop problems just before a crisis occurs.

Was there a strong sense of isolation at an earlier stage? Were there serious issues in school life? Were bullying, family problems, academic pressure, lack of sleep, or despair persisting for a long time?

Of course, the important thing here is not to link "people with mental health issues are dangerous."

You cannot simply associate mental distress with violence, and the majority of people with mental health issues do not harm others.

The discussion should not be about finding dangerous individuals.

It's about "finding people who need help sooner."

The "20-Second Screening" That Already Exists

When you think about understanding the state of the mind, you might imagine something like a large-scale psychiatric evaluation.

However, initial screenings are not necessarily complex.

The "Ask Suicide-Screening Questions (ASQ)" provided by the U.S. National Institute of Mental Health (NIMH) consists of four questions and can be conducted in about 20 seconds.

In NIMH's study, it was found that in a verification targeting young people aged 10 to 21, 97% of those at risk of suicide were identified if they answered affirmatively to one or more of the four items.

However, there is a very important caveat here.

A positive screening result does not mean a diagnosis of a mental disorder.

Nor should AI or schools decide that "this person is dangerous" or "has a high possibility of suicide."

It is merely an entry point that suggests "additional evaluation by a specialist may be needed."

This concept of "not a diagnosis but an entry point" is an important principle when utilizing AI in the mental health field.

AI's Strength Lies in "Changes Over Six Months" Rather Than "Today's Score"

The purpose of using AI is not just to have people answer one-off questions.

The greater potential lies in capturing patterns that change over time.

For example, suppose students conduct a simple check once a month for about three minutes.

For the first few months, there are no major issues.

However, gradually, sleep time decreases.

Next, the number of responses indicating "spending less time with friends" increases.

They begin to express strong anxiety about academics, and responses indicating "no hope for the future" increase.

Looking at each response alone may not necessarily indicate an emergency.

But when you line up the changes over six months, a clear downward trend may be visible.

It is difficult for humans to continuously track such subtle changes in thousands of students or employees.

AI can assist in this area.

In other words, AI does not become a "psychiatrist" but plays a role similar to a radar in weather forecasting.

It does not confirm abnormalities.

It alerts that "something different is happening than before."

In the Philippines, Screening in Schools Has Already Begun

Interestingly, in the Philippines, mental health screening itself is not entirely a future concept.

The Philippine Department of Education is advancing comprehensive health assessments for public school students through the Learners’ Health Assessment and Screening.

This includes mental health screening along with physical examinations, nutritional status, and dental check-ups.

For young people aged 10 to 19, age-appropriate screening methods are used, and it is assumed that they will be connected to specialized support if necessary.

Furthermore, the clinical guidelines for regular health check-ups in the Philippines also recommend screening for depression and anxiety in children and adolescents, and checking for anxiety, stress, and sleep disorders in adults.

With this in mind, the future discussion is not just about "whether to conduct mental health screenings."

It's about how much AI should be integrated into the existing system.

Who will see the results?

How will it be communicated to the individual?

To what extent will the data be stored?

And can a system be established to connect those who test positive to actual specialists?

These aspects of system design will become more important.

Voices on Social Media Wanting "Earlier Detection"

When looking at social media discussions on this theme, the expectation for early detection stands out.

 

Regarding the introduction of regular mental health screenings in schools, opinions on platforms like Reddit suggest that it could provide an opportunity for schools to grasp issues that students find difficult to discuss, such as family problems, abuse, severe anxiety, and depression.

The idea is that "children in distress may not always ask for help themselves."

There are also points appreciated by users regarding AI mental health support.

It is available 24/7.

There is no psychological barrier as high as talking to a human.

It helps organize confused emotions into text.

It allows you to vent worries you thought weren't worth consulting about.

This "ease of access" is one of the benefits that only AI can provide.

Especially in areas with a shortage of specialists or environments with strong prejudice against psychological counseling, digital tools may function as the first entry point.

On the Other Hand, "I Don't Want My Mind Monitored by Schools or Companies"

Equally strong on social media is the concern for privacy.

"Who will see the responses?"

"Will it be stored by the school?"

"If I'm judged to have anxiety or depression tendencies at work, won't it affect my performance evaluation?"

"Is there a guarantee that AI companies won't use the data for learning?"

These are the questions being raised.

Mental health information is even more challenging to handle than regular health data.

For example, if an employee answers "I feel strong anxiety,"

and that result is communicated to their immediate supervisor or HR, the individual may be less likely to answer honestly next time.

They might worry about its impact on promotions.

They might fear not being entrusted with important work.

They might be labeled as "mentally unstable."

It's natural to think this way.

On social media, similar concerns are repeatedly expressed about the records of AI-assisted therapy or counseling.

While some users appreciate the convenience, others react with "I don't know where extremely personal conversations are stored" and "I don't want AI as a third party in conversations with human therapists."

This is not merely a dislike of AI.

It's an issue related to the "environment where you can speak your mind safely," which is a prerequisite for mental health support.

The Most Dangerous Thing in the Workplace is "Health Support" Turning into "Personnel Evaluation"

When introducing AI screening in the workplace, this issue becomes even more critical.

If companies create a system where they can confirm individually,

"Person A has depressive tendencies,"

"Person B has high anxiety,"

"Person C is at high risk of burnout,"

it approaches employee monitoring rather than health support.

A more realistic approach is to separate personal responses from organizational information.

Only those with confidentiality obligations, such as medical and psychological professionals, can see personally identifiable information.

Meanwhile, only anonymized and aggregated information is shown to company management.

For example,

"Burnout-related indicators have significantly increased in a certain department over the past six months."

With this information, the company can consider organizational issues rather than identifying individuals.

Is overtime increasing?

Are the goals set unrealistic?

Is there a problem with management?

Is there a chronic shortage of staff?

Mental health is treated not as an individual's weakness but as a sensor reflecting the organizational environment.

In fact, on social media, there is much cynicism and distrust towards companies that advocate "well-being" or "employee mental health" while increasing the workload.

The suggestion that workplace environments causing stress should be improved before measuring employee stress with AI cannot be ignored.

What AI Should Not Be Allowed to Do

If AI is to be used in the mental health field, a clear boundary is necessary.

What should be particularly avoided is entrusting AI with "diagnosis," "prediction of dangerous individuals," and "decision-making on treatment."

AI's role is to detect changes and patterns.

From there, evaluation, diagnosis, counseling, and emergency intervention should be handled by qualified humans.

Additionally, systems that collect emails, chats, social media, and school behavior records without the individual's knowledge to estimate psychological states need to be considered with extreme caution.

If companies start judging "this employee may be in a depressive state" based on information not provided by the individual, it becomes a surveillance system under the guise of mental health measures.

The World Health Organization (WHO) also warns that while generative AI is being used as a psychological support partner by many, including young people, many of these systems are not designed or clinically validated for mental health purposes.

As AI capabilities rapidly improve, it's necessary to consider not just "whether it can be done" but "whether it should be done."

What We Need is Not an "AI Therapist" but an "Early Warning System for the Mind"

In this discussion, there's no need to imagine a future where AI replaces human psychological counselors.

In fact, the opposite is more realistic.

AI detects small changes and connects them to humans.

If within normal range, it provides information on self-care.

If the condition continues to deteriorate, it suggests a meeting with a counselor.

If high urgency responses are confirmed, professionals respond according to predetermined procedures.

That's sufficient.

To establish this system, at the very least, consent from the individual, minimization of data collection, strict access restrictions, clear data retention periods, supervision by professionals, and a mechanism for humans to reconfirm AI's judgments are essential.

Most importantly, there must be support available after the screening is conducted.

Even if AI detects 100 SOS signals, if there's only one professional available for consultation, the problem won't be solved.

##HTML_TAG_289