Could AI fight cyberbullying or ID suicidal depressives?

If you’re at all like me, it’s hard enough to keep a coupla digital gadgets working. Just ordered a new Mighty Mouse for my tired desktop and found I had to search among antiques on Ebay. My new refurbished buddy isn’t even a Magic Mouse, but does the trick.

So when it comes to artificial intelligence (AI), I’m at the mercy of doubters (no such thing yet) along the spectrum to the Cassandras (it’s already ending democracy, the world as we know it.) I DO suspect, that AI’s why I’ve had to teach my left pinky to hit tab to accept “spelled out” words as I’m typing or texting. Kinda hate it, BTW.

I knew a few students just a decade or so ago who thought AI would be a fun major – ya know, simple robot commands, chess-playing buddy, a way to get your phone or calculator to generate the best class schedule in college. Well, a few algorithms and coding/programming the first week, sent most running.

So, given my –and my bullied rescue dog’s- interest in reducing cyberbullying, our eyes and ears perked up a coupla months ago. Dean Chester’s article  discussed AI’s possible role in fighting cyberbullying.

Before getting to the article, an excerpt about AI in general – and in specific, about mental health (key cyberbullying issue) – from my fave NYTimes columnist,

David Brooks (June, 2019). His mention of A.L.S. in the first paragraph caught my attention since my sis-in-law battled that awful disease just a few years ago.

“Artificial intelligence is by turns terrifying, overhyped, hard to understand and just plain awesome.

For an example of the last, researchers at [U.C.S.F.] were able this year to hook people up to brain monitors and generate natural-sounding synthetic speech out of mere brain activity. The goal is to give people who have lost the ability to speak — because of a stroke, A.L.S., epilepsy or something else — the power to talk to others just by thinking.

That’s pretty awesome.

One area where A.I. can most immediately improve our lives may be in the area of mental health. Unlike many illnesses, there’s no simple physical test you can give someone to tell if he or she is suffering from depression.

Primary care physicians can be mediocre at recognizing if a patient is depressed, or at predicting who is about to become depressed. Many people contemplate suicide, but it is very hard to tell who is really serious about it. Most people don’t seek treatment until their illness is well advanced.

Using A.I., researchers can make better predictions about who is going to get depressed next week, and who is going to try to kill themselves.

The Crisis Text Line is a suicide prevention hotline in which people communicate through texting instead of phone calls. Using A.I. technology, the organization has analyzed more than 100 million texts it has received. The idea is to help counselors understand who is really in immediate need of emergency care.

You’d think that the people most in danger of harming themselves would be the ones who use words like “suicide” or “die” most often. In fact, a person who uses words like “ibuprofen” or “Advil” is 14 times more likely to need emergency services than a person who uses “suicide.” A person who uses the crying face emoticon is 11 times more likely to need an active rescue than a person who uses ‘suicide.’”

Can AI fight cyberbullying?

So, here’s   Dean Chester’s article  at Ben Dickson’s TechTalks from a coupla months ago.

[… It’s} apparent that the current levels of cyberbullying-awareness must be raised among people of all ages. There exist Internet safety guides for children as well as for parents and teachers, Stop Cyberbullying Day is recognized internationally and endorsed by profit and non-profit organizations, governmental programs work to help the victims, etc.

All of the above are necessary and are fighting a very important battle. But let us look at how we can use technology to help fight it.

Machine learning** against cyberbullying

Since technology is used in online bullying, it’s natural that technological solutions to the problem are sought.

Machine learning opens up a lot of possibilities to prevent cyberbullying. Currently, there are many initiatives to create and train algorithms that are able to detect hate and abusive speech online to block the user from seeing it and, therefore, getting cyberbullied.

The advantage such algorithms have over parental control software and keyword-spotting blockers is that they should recognize subtle and sarcastic comments—a task that the former solutions can’t cope with. Besides, the use of machine learning is necessitated because slurs and insults can often be, intentionally or not, misspelled.

In the paper Automatic detection of bullying in social media texts, a group of researchers describes such an algorithm. In its experimental state, it did considerably well recognizing abusive behavior online in English and Dutch. The scientists behind this project call the fact that their system can detect signals of bullying their main achievement.

This algorithm also determines who the bully, the victim, and the bystanders are in each situation, which can help a human moderator of a website do their job faster and more efficiently.

Another study brings the limitations of keyword filtering to closer attention. Based on data collected on Reddit, the researchers have arrived at the conclusion that many topical words are used by both hate groups and support groups, making it more difficult for a filter to determine which is which.

Instead, this study proposes to train an algorithm to recognize data produced by communities that “conform to linguistic identity” of hate groups. With that, it will be able to see patterns typical for such groups and communities in posts on social media and other Internet resources.

IBM Watson artificial intelligence is used by Identity Guard and Megan Meier Foundation to monitor children’s social media activity for signs of bullying or suicidal thoughts. If it sees any, it then reports such instances to the parents along with helpful resources aimed at dealing with the situation.

Cyberbullying causes a lot of problems for its victims but the most devastating one occurs when emotionally unstable people are targeted. As was the case with 12-year-old mentioned at the begining of this story, sometimes it can lead to self-harm and even suicide.

Machine learning algorithms can already identify suicidal tendencies with high accuracy based on “neural representation of emotions.” In the future, emotion AI can become an invaluable asset in recognizing and preventing self-harm. If a machine is capable of understanding human emotions based on voice or facial recognition, it will mean a tremendous leap forward for the use of AI in suicide prevention.

Despite their early stages being quite successful, AI anti-hate speech algorithms still have a long way to go. The biggest challenge for machine learning is probably how difficult it is to determine what is and is not hate speech. For example, the same sentence can be hateful and not depending on who writes it, their cultural and racial background, etc.

Besides, the definition of hate speech varies from community to community and from person to person. It can also change with time and new cultural norms emerging.

While most of the solutions mentioned above are still in the experimental stage, they all show a lot of promise. It’s important to remember, though, that even when they are properly implemented, it won’t mean the end of cyberbullying. On the contrary, something of an arms race can be expected between anti- and pro-bullying technologies.

So, it’s necessary to use AI for assistance but not expect it to do all the work.”