Basically where it could excel is giving probabilistic prediction of "mental disorders" across a high volume of users. What is a good thing you would do with that? Advertising to suggest someone to get help, maybe?
Our research group applied this tech to Belgium's suicide prevention center chat tool, so the workers can prioritize based on severity.
An appropriate use could be the treatment of those mental disorders on an 'industrial scale'. You can use the technology to identify as well as apply cost-effective tools that allow people with mental issues to heal and grow.
If you consider that "good" is another matter.
There's also the argument that software engineers may not always be in such high demand, and you never know what's going to be catalogued and indexed to you 10, 20, 30 years from now.
I was able to get by on working as a bartender while looking for a job. I had fun doing it and little cost of living.
Why should I have sacrificed my mental health or my happiness for a job working for assholes? (pardon my language)
I have no delusions that my political opinions or writing style are original enough or convincing enough to change the world. So it’s all just risk for me. If I’m writing something on my blog, it’s probably a focused technical article with no politics and little to no humor.
I actually despise these trigger warnings and cancel culture of cultural pieces just because the author was a misogynistic idiot (or anti semite, or proponent of slavery). We need to put bools/texts/essays into their historical contexts and discuss why we today hold beliefs that are different (and to us feel more modern/humane/better).
I am politically very left, atheist and I argue against dehumanizing isms since my days in high school 25 to 30 years ago. But still I find value in discussing ideas from Kant, Luther or many others that nowadays are often canceled in educational settings because of their despicable values/isms.
I mean, when it comes to stuff like mass-surveillance, we can debate the merits of having such a mass-surveillance infrastructure. But when it comes to simple calculatory tools, they're basically just thoughts -- thoughts that people can have even without AI, and presumably'll become increasingly well-developed over time.
Forward-facing ethical-frameworks should probably discount perceived scarcity of computation. This is, if we're talking ethics, presumably we ought to consider things under a presumption of ample cheap computing power, as that seems to be the direction we're going in.
In other words, it ought to become trivial for individuals to come up with better methodologies than this on a whim. And since it'll be so easy for most folks to think up (and implement) stuff like this on their own, there's little point in fretting over exploration of such things now. Rather, we presumably ought to regard stuff like this as mundane thinking.
We lend money for mortgages to people based on legitimate criteria: FICO score, income, etc.
THEN, we use a model to see if we lend more/less money to African Ameicans at better/worse rates. If we see bias in this model, we make adjustments to the first model so that they are "treated fairly".
I've seen the same done on message boards where certain actions were more likely for men vs women - and the administration discovered they were accidentally awarding more points to men vs women. So they adjusted the points for different actions so women had an fair-ish chance at getting points
Well, investigative officers concerned for the well-being of the individual and the potential for self harm or danger to others could attempt to help them with a taser or a shotgun.
Or, if you are a liberal you could have an algorithmic silo-ing of their posts? That way it would be demonstrably not a prejudiced censorship but something that could be investigated later if the right channels are followed.
It could be coupled with the software that detects whether people are homosexual or conservative based on their physiognomy.
the original claim was to trust no one. my comment was the government should also not be trusted.
people put on such a front on social i wouldn’t trust any analysis. especially not health related
> if the concern is abuse
not my intent, OP. the comment i replied to suggested gov instead. i made a comment about gov
there was as much reason to assume what i meant as much as what you assumed
population level data of social media is vulnerable to abuse by gov. social media is a poor input and gov will exploit ambiguity to their benefit.
1) The oversampling of the training set makes it uncalibrated when applied to the general population (to recalibrate you need an estimate of the prevalence to begin with, which sort of defeats the purpose).
2) Online posts are not a random sample of the population. (Perhaps this is solveable with some poststratification of the estimates, although requires demographic data on the poster.) If you take self reports that the researchers used to define disorders at face value, those would make more sense than using this model.
These text based models are so superficial, when applied to mass datasets with low prevalence of the underlying condition, they will ultimately result in very low positive predictive values (e.g. flag 100 people, if the model is good will only get 5/100 as actual mental health problems).
As version_five asks, it is hard to imagine any reasonable use of the model given such low positive predictive values (which imply incredibly high false positive rates).
this post sounds dangerous to me -- there are vast differences in a legal judgement and a medical judgement; thirdly, political policy levers are sometimes used in ways that make very little sense, due to situational factors.
It is no secret that humans are competitive, tend towards violence, and political oppression exists to some extent, of some groups, almost everywhere. A "medical judgement" used for legal or worse, political purposes, I claim, is dangerous. The post above this one appears to make no distinctions and leans towards "yes, do this."
Applying it to the statements of dictators so you know when to replace them?
We must apply this to spreadsheets ASAP.
this sounds like a fantastic tool to use psychiatric diagnosis the way it's supposed to be used: as a science-shaped cudgel with which to beat inconvenient members of society.
Remember that time in living memory when homosexuality was in the DSM. Now picture that but with machine learning able to pick out the gay posters online.