Link: https://www.koreatimes.co.kr/amp/southkorea/society/20260515/at-seouls-bridge-control-center-ai-helps-stop-99-of-suicide-attempts

Established in 2021, the center uses artificial intelligence (AI) for comprehensive emergency response, monitoring 900 CCTV cameras across 17 of Seoul's 21 pedestrian-accessible Han River bridges. Beyond suicide prevention, its most frequent task, the center also handles criminal tracking, traffic accidents and drug enforcement.

...

Much of that credit goes to AI, which triggers an alarm if an object identified as a person remains for more than 300 seconds in a bridge's "loitering zones," sections where people are able to stand for extended periods.

This is one of the good uses of AI. It is called object detection with neural networks and is a very classic use of convolutional neural networks (CNNs) in computer vision.
There was no LLM, no transformer, no huge data center necessary for training this model.

Please distinguish generative from predictive AI, it means a lot to all the data scientists out there inventing cool stuff!

I am not very uplifted to be honest since the same technology and devices can be used for large scale surveillance.

So instead of tackling the causes of suicide, they tackle the results.

So do the people attempting suicide get help after the ai detects them, otherwise wouldn't the person just attempt suicide somewhere else?

Huge middle finger to these people would probably be to get a fine for attempting suicide in the first place.

There's also the issue of adaptation. There's articles now explaining this technology which means if I'm in South Korea and I want to jump off a bridge I know I'll have to do it quickly or otherwise they'll come and get me. For some this might be the added pressure they need to go through with it.

Ultimately the best way to prevent suicide is to make life worth living and provide support for those who are in a mental health crisis. Neither of those things are going well in SK and AI surveillance won't fix that either.

good point.

Those people in the control center should keep an eye on multiple bridges across the Han river 24/7, with multiple sections for each bridge. I think it's a good thing to use AI in this case to reduce the human errors and weaknesses.

No, this is not a good use for AI. First, the concept here is to treat symptoms instead of the actual issue, second, stopping suicide attempts from happening isn't the only thing AI is used for here and third, in the context of second, a 15% hallucination rate means a lot of innocent people are harassed for doing nothing wrong.

Edit, because according to the votes and both answers many people seem to have misunderstand this: the 15% false positive means too many innocent people harassed when this surveillance system is used against crime.
On the suicide issues, the AI just doesn't solve the issue sustainably.

... triggers an alarm if an object identified as a person remains for more than 300 seconds in a bridge’s “loitering zones” ...

I don't know how the AI can hallucinates in such scenario, but it's better to harass some people to prevent some other people from committing suicide on those bridges.

I felt similarly. So I looked into what and how the country is handling this serious issue. The South Korean government has built a continuum of care, spanning federal initiatives down to community based programs.

National Policy, Service Delivery, Programs, and Data for Suicide Prevention in Korea

https://pmc.ncbi.nlm.nih.gov/articles/PMC12370428/

What the article is about: This 2025 research paper reviews how South Korea has developed policies, data systems, and community programs to address its suicide crisis — the highest rate among OECD (wealthy, developed) nations.

Why it's such a big problem: South Korea's suicide rate surged after the 1997 financial crisis and kept climbing during later economic downturns in 2003 and 2008. It peaked in 2011 at 31.7 deaths per 100,000 people. Since then it has declined overall, but suicide rates among people in their 20s have been trending back upward — a concern the paper flags as needing new solutions.

How the government has responded: Beginning in 2004, the government launched a series of five-year action plans. In 2011, they passed a formal Suicide Prevention Act, giving these efforts legal structure. The current plan (2023–2027) aims to cut the suicide rate by 30%.

Who is most at risk: Beyond young adults in their 20s, the data shows rising suicide attempt rates among people under 20, particularly females. People living alone, those with lower incomes, and individuals diagnosed with conditions like depression, bipolar disorder, schizophrenia, alcohol dependence, and insomnia are also identified as high-risk groups.

What they're actually doing: Training millions of everyday people to recognize warning signs in those around them — over 7 million trained as of 2022
Expanding mental health screenings and shortening how often adults are screened (from every 10 years to every 2 years as of 2025)
A pilot program connecting regular doctors' offices to mental health resources, since the paper notes that people at risk are more likely to visit a general doctor than a mental health professional before a crisis
Offering vouchers for up to eight sessions of free professional counseling
Analyzing police records and health insurance data together to better understand who is dying and why

What the article says isn't working: Local governments often lack the staff, funding, and infrastructure to actually carry out what the national government plans. Previous action plans also tended to react after crises rather than prevent them proactively. The paper also notes that suicide prevention has focused too heavily on individual mental health while not doing enough to address broader social and economic factors — things like job instability, economic hardship, and the specific pressures facing young people and women in Korea's labor market.

National Trends and Directions for Suicide Prevention Research in South Korea: A Narrative Review

https://psychiatryinvestigation.org/journal/view.php?number=1905&viewtype=pubreader

Summary: What this article is about: This is a 2025 review paper that looks at research on suicide prevention in South Korea, covering what strategies have worked, what the risk factors are, how high-risk individuals are being managed, and where digital technology fits in.

The scale of the problem: South Korea's age-standardized suicide rate of 23.6 per 100,000 is more than double the OECD average of 11.1, making it the highest among OECD countries. During COVID-19 (2020–2022), deaths from suicide were actually 26% higher than deaths from COVID-19 itself. Beyond the human toll, a 2015 economic analysis found that suicide ranked first among all diseases and injuries in terms of financial burden on the country, estimated at $8.3 billion — the largest single burden for people in their 20s through 40s.

What has actually worked:
Restricting access to lethal means has had measurable results. After a ban on the pesticide paraquat in 2011, suicide deaths from pesticide poisoning dropped by nearly half — from 5.26 to 2.67 per 100,000 — between 2011 and 2013. Installing full-height platform screen doors at 121 Seoul subway stations between 2005 and 2009 reduced fatal subway suicides by 89%, from roughly 100 cases to 11.

Media guidelines have also made a difference. After celebrity suicides, daily suicide deaths increased by an average of 6.27 per day in the following 30 days. After stricter media reporting guidelines were put in place in 2013, the suicide rate declined by 13%. The guidelines — now in their third version — prohibit detailed descriptions of methods, sensationalized headlines, and require crisis helpline information to be included in any suicide-related reporting. Monitoring of social media content is now part of these guidelines as well.

Gatekeeper training through a program called "Suicide CARE" has trained over 8.6 million people since 2013. Trained gatekeepers showed improved intervention skills and confidence, though the article notes more research is needed to measure long-term impact on actual suicide rates rather than just awareness.

Managing high-risk individuals: People who have previously attempted suicide are at high risk of reattempting within 24 months. Since 2011, case management teams in hospitals have supported suicide attempt survivors in emergency departments. A study found those who completed four weeks of case management had a significantly longer time before any reattempt. Family members of those who have died by suicide are also identified as high-risk — their suicide rate was found to be 586 per 100,000, three times higher than bereaved families of non-suicide deaths.

Digital tools: South Korea has a 95% smartphone penetration rate, which the article identifies as an opportunity. Mobile apps for safety planning, AI-based chatbot counseling, and mobile messenger follow-up for people discharged from emergency departments have all shown early promise. Machine learning is also being explored to predict suicide risk. However, the article notes these are still in early stages and need more research.

What's still falling short: Korea's suicide prevention budget, while growing, remains roughly 1/20th the size of Japan's. The article argues South Korea needs to go further — establishing dedicated clinical research centers, expanding hospital-based case management, and developing AI-assisted tools — pointing to Denmark and Japan as models where similar investments led to measurable reductions of 22–24% in suicide attempts.

Korea sets ambitious goal of cutting suicide rate by 40% in 10 years

https://www.koreatimes.co.kr/southkorea/health/20250912/korea-sets-ambitious-goal-of-cutting-suicide-rate-by-nearly-half-in-10-years

What this article is about: A September 12, 2025 news report from the Korea Times covering the South Korean government's newly announced national suicide prevention strategy.
The current situation: South Korea has had the highest suicide rate among OECD countries for 22 consecutive years. In 2024, more than 14,400 Koreans died by suicide — an average of 40 per day. The current rate stands at 28.3 per 100,000 people.
The goals: The government wants to cut annual suicide deaths to below 10,000 within five years, and reduce the rate from 28.3 to 17 per 100,000 by 2034. Officials explained they chose 17 as the target by referencing Lithuania's rate of 17.1, which is the second-highest among OECD nations.

What the government plans to do:
Emergency room data from suicide attempts will now be automatically shared with local suicide prevention centers for immediate follow-up. Previously, local officials were only notified if police or fire departments requested it and the person consented.
The number of suicide risk response centers will expand from 92 to 98 next year. These centers provide counseling, psychological assessment, and economic support to high-risk individuals, including family members of those who have died by suicide.
To address financial hardship as a crisis trigger, the government plans to purchase and write off long-term overdue loans held by small business owners and individuals overdue for at least seven years and worth less than 50 million won (about $36,000).
The Ministry of Education is strengthening school violence prevention and support for victims. The Ministry of Employment and Labor is intensifying efforts against workplace bullying, including stricter labor inspections and penalties for firms where serious incidents occur.

The health ministry also plans to revise laws to better support people struggling with addiction. Two additional national suicide hotline centers will open next year. AI tools for counseling analysis and online risk monitoring will be used to improve detection.

What officials said: The second vice minister of health acknowledged there were reservations about setting a specific numerical goal, but said simply pledging effort without a target was not enough. He added that existing measures can still yield better results depending on the commitment and urgency with which local and central governments apply them. Separately, 115 legislators proposed a resolution declaring the government responsible for keeping citizens "safe and happy" and pledging to work toward a significant reduction in the suicide rate.

The Role and Future Tasks of the National Assembly Suicide Prevention Forum - Focusing on the Legislative Content of Suicide Prevention Laws

https://pmc.ncbi.nlm.nih.gov/articles/PMC11298267/

What this article is about: A 2024 paper published in Psychiatry Investigation examining the history, accomplishments, and future direction of South Korea's National Assembly Suicide Prevention Forum — a group of lawmakers dedicated to advancing suicide prevention through legislation and budget allocation.

The scale of the problem: In 2022, there were 12,906 suicides in South Korea — 4.7 times the number of traffic accident deaths that same year. From 2020 to 2022, suicide deaths totaled 39,453, exceeding the approximately 35,000 COVID-19 deaths during the same period. Suicide is the fifth leading cause of death overall, and the leading cause of death for people in their teens, twenties, and thirties. The rate climbs sharply with age, reaching 60.6 per 100,000 among those aged 80 and above. Roughly 90,000 people are added to the bereaved family high-risk group each year, and the annual socioeconomic loss from suicide is estimated at 5.4 trillion Korean won.

Three turning points in South Korea's suicide prevention history: The article identifies three key milestones. The first was the establishment of the Korea Association for Suicide Prevention in 2003–2004, a private nonprofit bringing together experts from medicine, media, law, and local communities. The second was the passage of the Suicide Prevention Act in 2011, which formally acknowledged suicide as a serious social harm and established the national duty to prevent it. The third was the creation of the National Assembly Suicide Prevention Forum in 2018, formed by 39 lawmakers, which the article describes as completing a prevention network by adding legislative and parliamentary power to what civil society and government had already built.

What the Forum has done: Since 2018, the Forum has held 23 policy seminars covering causes of suicide (13%), policy directions (17%), target groups such as workers and adolescents (22%), budget (9%), organizational structure (9%), and other topics (30%). It has also held six international seminars, bringing in experts from the WHO, Australia, Denmark, and the United States to share practices from countries that had already reduced their suicide rates. The Korean Neuropsychiatric Association played a central role, chairing 11 of the 23 seminars and presenting at 13. Following the first seminar, the Ministry of Health and Welfare began releasing provisional monthly suicide statistics starting in 2020 — something that had not been done before. Budget increases were also directly tied to Forum activities: the suicide prevention budget rose from 7.3 billion won in 2017 to 13.6 billion won in 2018, and continued climbing to 48.9 billion won by 2023 — a 6.7-fold increase. The article notes, however, that this still amounts to roughly 1/20th of Japan's suicide prevention budget of 830 billion won.

Legislative achievements: In its first term (2018–2020), the Forum proposed 17 bills, four of which passed. Key outcomes included establishing the Suicide Prevention Policy Committee under the Prime Minister's Office, requiring suicide reporting guidelines to be included in the national five-year plan, enabling information sharing between police, fire departments, and suicide prevention centers, allowing emergency services to request location data for individuals at suicide risk, prohibiting the distribution of suicide-inducing information, and adding bereaved family support to the responsibilities of suicide prevention centers.
In its second term (2020–2024), 14 bills were proposed and three passed. These included allowing suicide prevention centers to obtain personal information on suicide attempters without their consent — shifting from a passive to a proactive approach — formalizing the legal foundation of the Korea Foundation for Suicide Prevention, and mandating suicide prevention education across national agencies, schools, workplaces with 30 or more employees, and senior welfare facilities, effective July 12, 2024.

What still needs to happen: The article argues that legislation alone cannot prevent suicide and that the Suicide Prevention Act by itself is insufficient. It calls for a broader review of other laws that affect suicide risk, the creation of a dedicated Suicide Prevention Fund for stable long-term funding, and a continued shift toward addressing the underlying social and economic conditions that drive people toward suicide rather than focusing only on individual mental health responses.

I've always read that suicide is usually an impulse decision, and if you can block the attempt, many people won't try again.

https://www.npr.org/2008/07/08/92319314/in-suicide-prevention-its-method-not-madness

Why is AI necessary for that? I'm not a programmer, but that seems like a simple if/then statement

Sure, but you need a way to identify whether an object is a person. That's a lot less simple.

That's interesting. From what I understand people who commit self harm often do it multiple times, so I assumed it was the same for suicide

When it comes to suicide prevention, I'd prefer having a few false positives than any false negatives.

Stopping people from dying is good.

Still don't need AI for that

Still, the technology has its weaknesses. Kim said the system carries a hallucination rate of about 15 percent, including instances where it misidentifies an object as a person, which is why human judgment remains the final call.

That is how it hallucinates.

Besides that, you unfortunately also did not read my comment completely. I specifically pointed out the other instances this AI surveillance system is used on, the article brings up crime as an example. That is where 15 percent hallucination rate means a damn lot of innocent people get harassed by police.

That's not the issue where I criticised the false positive rate though, if you read my comment carefully. I hope I made it clearer with my edit.

Thanks for the two pages of text, very helpful. Can you tell me what you just said without wasting more tokens? This is not "looking into it", this is wasting resources to produce an unreasonably long text that may or may not be true

triggers an alarm if an object identified as a person remains for more than 300 seconds in a bridge's "loitering Zones"

Is this one of those things where we're just calling everything AI? This seems like a script with machine learning object recognition, and definitely not a determination you'd want to leave up to an AI.

Oops, my bad for didn't read the whole article.

Still, the technology has its weaknesses. Kim said the system carries a hallucination rate of about 15 percent, including instances where it misidentifies an object as a person, which is why human judgment remains the final call.

The AI only flagged the people (or the objects it misidentified as people), but the human still decides whether those people are worth checking on. I think it still the human's fault if a lot of innocent people get harrassed by the police.

I guess it would be AI or machine learning specifically with OpenCV and attach it to a pre trained model. I guess that could work and slap on AI for increased price

See? AI saves lifes. Now go pend your tokens and stop using your brain, meatbag /s

Still, the technology has its weaknesses. Kim said the system carries a hallucination rate of about 15 percent, including instances where it misidentifies an object as a person, which is why human judgment remains the final call.

Don't worry, they treat it as a tool, like how people treat doorbell cameras with motion detection.

This system totally makes sense considering many of its citizens call the place Hell Joseon.

Did you get from the article that they're not also doing that, or is that from your own knowledge of the mental health care in that country?

It's why when the UK brought limits on the amount of Paracetamol you could buy in a single transaction in the early 00s the amount of self positioning from it dropped massively. You would think that people would just go to the next shop down the road or come back later in the day.

Damn, I cant solve societal problems, might as well just let them jump!

Youre a fucking moron.

Well, if you choose to misread my comment like that, AI can't stop you.

Which they mention in the article:

The center's reach extends beyond rescue. The system supports filtered searches by gender, age and clothing type — a search for "April 29, male, Mapo," for example, pulls up footage of every adult male who crossed Mapo Bridge that day, helping police map the movement routes of suspects. All data is strictly managed under the Personal Information Protection Act and deleted after one month.

Emphasis mine. Who thinks they actually delete those records?

You can do both. Also a lot of people who try to commit suicide don't try immediately after they fail, so picking them up alive gives you time to help them. If it was like "well whatever, we'll pick them up on the bridge so we don't have to worry about tackling the causes of suicide" i'd agree with you 100%, but that is not what this news is about. It's about improving a method to prevent the suicide from happening, like nets around high buildings. It's sad the root causes aren't fixed with this, but it's still a good thing happening imo.

Oh no someone might ask the victim of a false positive if they’re okay. The horror.

RIP birdwatchers and photographers, haha.

I guess you never have been targeted by unfounded police action. I hope you never will. It is fucking scary and traumatising.

American police is in its own category when it comes to cruelty in “police action”.

Don’t generalise it to other countries where becoming a police officer isn’t a 3 week online course.

There are very few things we as humans “need”.

But a ton of things that make stuff easier. Like using “AI” to detect humans in this case

Or don’t go to a bridge?

Bridge jumpers aren’t usually the ones that just walk off the edge with zero hesitation.

I am not talking american police. I'm talking about experience with german police. You know, where you have to go to university for three years.
The training doesn't matter when several armed officers are applying forceful measures against you, you have no idea why, are panicing and your panic reaction is read as resisting police officers. Because that is standard procedure.

I certainly agree this does not address the root issue, and it does normalize this technology for adresssing social issues and can then be applied to places where its more harmful like crime- And also I think taken on its own merits this is a good intervention to exist for suicide prevention specifically (though I think its valid to not take it purely on its own merits if one is concerned about the adoption of surveillance tech)

I do want structural solutions though. But that requires rebuilding society in the image of treating humans better, better work and environmental conditions, rethinking healthcare (at least here in the US, no idea how things are in Korea), and about a million other things

And I do think the normalization of surveillance infrastructure should be taken seriously. Especially with the present global rise of authoritarianism

That's a good point. Good thing the area is very highly surveilled and recorded.

Good day.

I'd like to preface that I'm replying in good faith.

I can to a degree, agree that the summarized wall of text that I provided, wasn't the best and could have falsehoods in it. However, being that I am not writing a thesis on the matter, applied vetting to sources, figured that anyone who reads this on this forum would apply at least a grain of salt to it, and that I wanted to help shed light on the good/better things that are going on... I figured it would be alright to post.

As for the the use of AI to summarize... (checks notes) ... over 11,000 to ~13,000 words (with notes and references,) down to ~1,700 words, I'd say it's helped us. And yes, these LLM's can be wrong, partially correct and even some cases, hallucinate. And that's why you shouldn't take all that I posted above as pure fact. But don't we live in a messy world? Haven't we as humans learned to pick out the lies, falsehoods, half-truths, and apply salt to statements and claims when we see it appropriate? It's not an excellent tool for all things. But it sure would beat me in summarizing it for everyone. Openly, my ADHD would have had me considering to give up a lot sooner if I had to do the summarizing myself. And furthering the openness, I considered a few times to not even work on replying with my findings.

So were the 'tokens' used a genuine waste? I would disagree. Do AI server farms cost a lot to operate and use an enormous amount of resources? Absolutely! https://www.technologyreview.com/2025/05/20/1116327/ai-energy-usage-climate-footprint-big-tech/ But as for my take on the use of conventional AI, this is one of the main reasons I was looking forward to the early advancements. (Like mentioned above, ADHD makes it harder to sit through a wall of text to read. Even if it's a subject or topic I'm really into.) And I was also looking forward to these advancements and positive changes. https://lemmy.world/post/46231655?scrollToComments=true

people only said suicide detection was a good use of AI, not crime surveillance. and nobody's pretending stopping a suicide attempt treats the underlying issues either, and that's still better than not stopping it

machine learning is AI

Yes you do, because an object on camera can look a lot of different ways. This is very different from LLMs like ChatGPT. It trains on different images like a Captcha and gets positive reinforcement if it identifies something correctly and negative if it is incorrect.

Machine Learning like this has been in use since the days of early digital computing. There isn't a more efficient way to achieve something like this.

They do something similar in the US for crimes. Why prevent crimes with welfare, single payer health care and a living wage when you can arm the police to the teeth with tanks and increase prison space.

Wow, you've convinced me

Just extrapolating from personal experience