Data Mining/Algorithms and Sanctioned Suicide

[XiaroX]

Breaking the silence: leveraging social interaction data to identify high-risk suicide users online using network analysis and machine learning - Scientific Reports

Suicidal thought and behavior (STB) is highly stigmatized and taboo. Prone to censorship, yet pervasive online, STB risk detection may be improved through development of uniquely insightful digital markers. Focusing on Sanctioned Suicide, an online pro-choice suicide forum, this work derived 17...

www.nature.com

Read this article. It demonstrates some of the ways what we post here is being used. What do you think? Will it ultimately help, or is it just intellectual posturing that ignores the deepest issues in life so that some guy can earn praise and degrees without crediting people like us? Does it make you feel dehumanized, or hopeful? Is it a good thing to focus on who is at most risk for suicide, to prevent suicide, or is it better to understand why people don't agree to Life's Terms and Conditions?

I am really hoping people respond to this thread. I can understand if reading the article seems like too much, often I can't concentrate, or just don't have the energy, and if no one responds, I'm going to try to find it in me to rip this guy to shreds. Fuck your fucking algorithms.

 

[U. A.]

You need -1 more posts to view this content

 

[Forever Sleep]

Truthfully, I didn't read the article- sorry. A combination of lazyness and being nervous to click on off site links...

I guess it depends on how they are using the data. If it's to better understand suicidal people to create better support systems for them. Maybe to educate therapists etc. that there are many more than one type of suicidal person then- that sounds positive.

If the sole focus is suicide prevention though- by proving that surveilling people's posts can give an accurate prediction as to whether they will or will not CTB- that seems incredibly intrusive.

Plus- I question the accuracy too. Some members simply leave the site. Doesn't necessarily mean they killed themselves. A goodbye thread doesn't necessarily mean soneone went through with the attempt. Being (necessarily) annonymous, I question the accuracy of relating a certain 'style' of thread to someone who actually will or won't attempt. Seeing as they can't be sure on whether we have in fact attempted always.

If all they're going to do is put this forward as a legitimate reason to spy on people even more. Intervene in their lives to stop them trying to escape- without actually even bothering to figure out why they are so unhappy to begin with and, focus on that. Then, this just seems a continuation of the jail system we already have in place. Nevermind how shitty their lives are! Just keep them living them! Thanks a bunch.

The above excerpt from U.A also seems pretty shallow in its interpretation. Yes- there will be members here who aren't convinced they will kill themselves. Yes, many of us do love it here for the sense of community. To be able to express things we would receive intense hostility and horror to everywhere else.

Some of us linger because we are waiting for other family members to die first though. I wonder if their AI model or they would look at my posts and the length of time I've been here and decide- they're not serious about it. Ultimately, time will tell. Hopefully one day, I'll get to prove them wrong!

I kind of doubt there is an algorithm as to predicting who will or won't suicide. Truthfully- some people here I had a feeling wouldn't seriously attempt- have. Whether they succeeded or not. I don't think you can always tell. Even on a site where people are likely expressing themselves fairly openly.

Ultimately, suicide can be a moments work. If the thougts are already there, who can truly predict what people will do moment to moment? That's surely why these 'helplines' focus on whether someone does have a method prepared. They don't really want to risk whether they will or won't use it. I'm not sure a computer can accurately predict the human mind.

 

[itsgone2]

First, I tried reading it. His point in my life many years ago, where I would've been interested in trying to figure out some of this academic components to it, but not now. My brain is too fried. Too much insomnia too many battles too much suicidal ideation.
So I tried and failed. I skimmed at some is the best I could do. So with that being said, it's looking through the site. Or it did anyway. It seems to be they're trying to find ways for prevention. Then I guess they're smart to do it. The person like me I guess you could analyze post and try to figure out how to stomp attempts. But why? Things in my life and in the world that led me here are still true. If those facts were different, I wouldn't be suicidal.
To be honest, it's the opposite. We all know it. It's discussed here frequently. They should be developing algorithms to figure out who should be allowed to peacefully go. We'll say Nembutal.

If they were doing great research, they'd be helping us escape. Not finding new ways to trap us here.

 

[heywey]

I'm not super familiar with this space, but the methodology seems iffy. Only 192 users, only 48 of those HRUs (aka people who committed suicide). Another problem I see is in how they selected HRUs, there's going to be a lot of bias since they filtered to a small number of commonly-used goodbye post keywords, and a lot of SaSu specific ones on top of that, where there's a lot more variety in reality. (I'd be very curious about the distribution of goodbye keywords in the HRUs they selected, but they don't say)

The approach is interesting though -- the model isn't actually trained on any post contents, only things like number of interactions, who interacted with who, and so on. I'm surprised it works as well as it does honestly. But at the same time I'm kind of not? I guess that's sort of my issue with it in general -- all the findings are obvious and easily explainable to anyone who spends more than 30 seconds looking at this forum and considering how things would go, but at the same time, they're not generalizeable in the slightest. Congrats, you can algorithmically figure out who's likely to commit suicide... on a suicide forum...

I don't know, I'm not inherently against people using SaSu data for things, but this just seems completely pointless to the point of tastelessness. If it could actually do anything useful elsewhere or provide insights into this forum's dynamics that'd be one thing, but to use SaSu data for a toy project like this is dumb at best and ghoulish at worst.