Facebook deploys AI for early signs of suicide • Hacker News Board

Re: Facebook deploys AI for early signs of suicide

#131

Post by AlexandrB » Tue, Nov 28, 2017, 3:27 AM UTC

Shank wrote: post #78

There's a lot of pessimism in the comments. I believe this is one of the best uses for an AI that we've come up with. First, let's go into the details: 1. They only monitor and elevate to moderators content from posts and Facebook live. 2. Moderators can then take action, and reports from this system are prioritized over other non-pressing issues. The actual life saving chances of this system are huge. Just from a pu…

> But if we're going to use AI in those fields anyway, we may as well extend it to trying to save lives.

Absolutely, but the incentives of the organization that does so need to be aligned with the people it's trying to help. Facebook's incentives are aligned with those of their advertisers not their users.

Even if this system is being conceived by well-meaning people with the best intentions the institutional pressure will be to monetize any information gathered and any conclusions arrived at from that information. It won't happen right away, but it's inevitable. For a contemporary example look at Google's transformation from "Don't be evil" to "we're going to gather tracking data on you wherever and whenever we can".

Re: Facebook deploys AI for early signs of suicide

#132

Post by BurningFrog » Tue, Nov 28, 2017, 3:36 AM UTC

asdgkknio wrote: post #113

flashman wrote: post #100

Earlier quoted context omitted.

> Facebook thinks they have the ability and the right to diagnose their users with mental illnesses Maybe it's more that Facebook thinks it's horrible when people kill themselves, and that if Facebook's tech could help prevent suicide, inaction would be immoral. Most of us accept some level of State interference in our lives for our own good. Maybe we are going to have to have that discussion about involuntary interv…

>Maybe it's more that Facebook thinks it's horrible when people kill themselves, and that if Facebook's tech could help prevent suicide, inaction would be immoral. That doesn't entirely contradict what I said. We generally assume psychologists and therapists are benevolent, but they aren't allowed to disclose private information even if it's in the best interests of the patient. Even if we assume Facebook is perfectl…

> They've thrown out all established codes of ethics

They may have nudged on one particular code of ethics.

Alarmism does not help your credibility.

Re: Facebook deploys AI for early signs of suicide

#133

Post by tw1010 » Tue, Nov 28, 2017, 3:36 AM UTC

asdgkknio wrote: post #15

This may or may not be a useful technology, but the fact that Facebook thinks they have the ability and the right to diagnose their users with mental illnesses is disturbing. They have more information about their users than psychologists have about their patients. They can (and do) build a psychological profile and diagnose mental illness. Yet rather than keeping this information in the closest confidence, they sell…

And does it not seem likely that the output of this model would be used as a signal to influence the newsfeed algorithm? It's like a feedback machine showing already depressed people pictures of happy families, making them even more depressed, yet more hooked, eventually leading to you-know-what.

Re: Facebook deploys AI for early signs of suicide

#134

Post by rootsudo » Tue, Nov 28, 2017, 3:37 AM UTC

anthrowaway wrote: post #129

anigbrowl wrote: post #107

Earlier quoted context omitted.

I help to run a suicide/depression support group on FB and I do not at all trust FB as a corporation to do a good job on this. Having had friends who've been involuntarily hospitalized and all that goes with that I'm extremely skeptical that this will turn out well. 'But it might save lives!' is true, but it might also wreck lives by institutionalizing people that thought they were venting their feelings to a trusted…

For those who aren't familiar, this is how it can play out: 1. Friend^H^H^H^H^H Facebook notices you said something that sounds suicidal to them. 2. They call the police. 3. The police show up and take you to an emergency psychiatric facility (you don't have as much choice in the matter as you think you do, and you can't anticipate the consequences). 4. The facility keeps you in their custody until your friends or re…

Agree 100%. Happened to me too when I went to the ER for a panic attack.

I reported the doctor and he told me to later go fuck myself.

What a world. I got baker acted for going to the ER for a panic attack because of my anxiety.

Re: Facebook deploys AI for early signs of suicide

#135

Post by AlexandrB » Tue, Nov 28, 2017, 3:38 AM UTC

tareqak wrote: post #98

Please permit me to play the pessimistic man on the street: If Facebook can deploy AI to detect the early signs of suicide, why can't they deploy AI to detect "fake news" or those susceptible to "fake news"? Furthermore, "cui bono" (for whose benefit [0])? What does Facebook gain from being able perform this sort of detection given it is a for-profit corporation and not a non-profit philanthropic organization? An asi…

> Please permit me to play the pessimistic man on the street: If Facebook can deploy AI to detect the early signs of suicide, why can't they deploy AI to detect "fake news" or those susceptible to "fake news"?

It's not about detection success rates, but economic outcomes if they screw up. If Facebook claims to detect fake news, but does it incorrectly they will be pilloried in the press (again). If they detect suicidal behaviour wrong the victim of said detection has little recourse.

In short, it's far safer to use vulnerable individuals as guinea pigs for AI than organizations that buy (digital) ink by the barrel.

Re: Facebook deploys AI for early signs of suicide

#136

Post by ThrowawayP » Tue, Nov 28, 2017, 3:45 AM UTC

Shank wrote: post #78

There's a lot of pessimism in the comments. I believe this is one of the best uses for an AI that we've come up with. First, let's go into the details: 1. They only monitor and elevate to moderators content from posts and Facebook live. 2. Moderators can then take action, and reports from this system are prioritized over other non-pressing issues. The actual life saving chances of this system are huge. Just from a pu…

Bear in mind that what will occur is that armed police will be dispatched to take the unfortunate individual into custody and transport them to a hospital to be involuntarily locked up in the mental health ward for psychiatric evaluation and it can be some number of days before they get a judicial hearing to attempt to secure their own release. (See https://en.wikipedia.org/wiki/Involuntary_commitment_interna...) By all reports, it can be a highly traumatic and humiliating process.

All this, based on some algorithm of unknown quality flagging posts for evaluation and reporting by moderators with no psychiatric or other medical training. And somehow this is supposed to be a good thing?

Re: Facebook deploys AI for early signs of suicide

#137

Post by AlexandrB » Tue, Nov 28, 2017, 3:46 AM UTC

foobaw wrote: post #32

There's a lot of people who are fearful that Facebook probably knows too much about its users and that could be problematic. But what do you expect? How do you want our future to be shaped? Innovation, including AI, requires tremendous amount of data. Suicide is a serious problem and being able to accurately identify that in the future could save tons of lives. Accurately diagnosing mental illnesses will take a while…

> But if we assume that Facebook is acting purely for the benefit of the society and the people, I think this is a great step.

Why would anyone ever assume this? Does anything in Facebook's past suggests that this might be the case? And even if the motives of the individuals who spearheaded this were pure is there any reason to think that this won't change in the future as people at Facebook change roles or move on?

Re: Facebook deploys AI for early signs of suicide

#138

Post by sjg007 » Tue, Nov 28, 2017, 3:52 AM UTC

dsp1234 wrote: post #95

sjg007 wrote: post #67

Earlier quoted context omitted.

I think it is unclear. For example Apple has HealthKit which collects health related data.. That data is probably HIPAA protected. If it's not (and if Facebook is not) then I would imagine we will see legislation and/or lawsuits to clarify things.

Apple is not a covered entity under HIPAA[0], and thus has no liability for the data it collects. "If an entity does not meet the definition of a covered entity or business associate, it does not have to comply with the HIPAA Rules." [0] - https://www.hhs.gov/hipaa/for-professionals/covered-entities...

Apple is most likely a business associate by definition.

Re: Facebook deploys AI for early signs of suicide

#139

Post by MBCook » Tue, Nov 28, 2017, 3:53 AM UTC

cryoshon wrote: post #62

jimmytucson wrote: post #20

Wonderful example of how we can use AI for good. Now, if this catches on, I can see Facebook adapting their algorithm to recommend treatments for depression, anxiety, even ADHD. This would be a huge success. "Mr. Smith, I see Facebook's AI considers you a strong candidate for medication X and your doctor agrees. Can you explain then why you haven't been taking medication X?" Also, if an algorithm can detect when a pe…

>I can see Facebook adapting their algorithm to recommend treatments for depression, anxiety, even ADHD imagine this: employers know that having these conditions (some of which are permanent) makes people less effective at working, and so they want to know who has adhd, who has x, y, z. facebook sells them the data. boom, now these people can't be employed by that company. i just can't help but laugh because there's…

Of course Facebook couldn’t sell that data. It’s medical information and thus illegal (IANAL).

But no matter. I’m sure there’s no rule against the company taking out ads advertising open positions to people who don’t have those issues.

Tada. Nothing illegal happened, but discrimination still did.

Re: Facebook deploys AI for early signs of suicide

#140

Post by ProAm » Tue, Nov 28, 2017, 3:54 AM UTC

jacquesm wrote: post #116

ProAm wrote: post #72

Earlier quoted context omitted.

Great points, I wonder if FB actually employees psychologists and if so how many and what are their accreditations and level of professional experience.

Yes they do. Main mission: to make Facebook as sticky as possible.

Is there proof though? Id assume so, Im not sure psychologists follow a Hippocratic oath if they did this seems like it would cross a bit of a line, no?