> Moderation wouldn’t really be fast enough to stop that
Social media keep using this excuse for not trying. We can moderate spam in emails with a simple naive bayes classifier, why don't we just do that with comments? It could easily classify comments that are part of a bandwagon and flag them automaticly hiding them or for human review.
We are able to moderate email but the concepts we use to do so are never applied to comments, I don't know why, this seems like a solved problem.
If you're trying to use the SPAM model as some kind of example of success I believe you may have already failed.
In SMTP servers I've managed for clients we typically block anywhere from 80 to 99.999% (yes 10000 blocked to one success) messages. I'd call that MegaModeration if there was such a term.
And if you think email spam is solved then I don't believe you read HN often as there is a common complaint of "Gmail is blocking anything I send, I'm a low volume non-commercial sender"
In addition email filtering is extremely slow to react to new methods, generally taking hours depending on the reporting system.
Lastly, you've not thought about the problem much. How are you going to rapidly detect the difference between a fun meme that spreads virally versus an attack against an individual. Far more often you're going to be blocking something that's not a bad thing.
Fair concerns but I have trained a Naive bayes classifier on twitter data in the past using [1] a social study of categorised tweet to train the classifier and got around 85% accuracy. It was able to detect and properly classify rape threats as abusive but conversations about rape seed oil as non abusive. Considering the small data set and how little entropy there is between samples I consider it pretty useful.
I get that no machine learning is 100% perfect which is why it should be used as an indicator rather than the deciding factor.
I have had issues with gmail blocking emails but as you point out it was always because of ip reputation not over zealous Naive Bayes.
Training classifiers can also go off the rails under adversarial attack. This commonly showed up in our systems when people sent short emails that were more ambiguous. For example this tends to cause problems where malevolent users adopt dogwhistles co-opting the language of the attacked group. The attacked group commonly becomes the ones getting banned/blocked in these cases
Okay, I actually did laugh out loud a little at the ‘we are able to moderate email’, bit.
Spam filters are probably one of the single most consistently unreliable pieces of software I ever have to use; regardless of the email provider; or email client I use.
I have to check my junk folder like it’s my inbox.
On both Apple Mail and Outlook; with two different emails - email money transfers (EMTs) will get shoved in my junk box; despite the dozens of times I have marked said emails as not junk.
I’ll get spam emails, but I don’t get mail from newsletters I’ve actually signed up for.
Like…if you’re trying to use spam emails as an example of success; and even a model we should follow for…anything else; I’m going to laugh you out of the room and tell you to keep me the hell away from whatever tools you want to use with that technology.
Spam filtering software for email is at best useless; at its worst; mind numbing log frustrating. It’s a tool I’ll never trust.
That's the current system. ML plus humans to remove harmful content. And people like Elon are extremely upset about this. Heck, you even see the GOP complaining about spam filtering on gmail. Hard to say that this is a solved problem that everybody agrees works well.
Social media keep using this excuse for not trying. We can moderate spam in emails with a simple naive bayes classifier, why don't we just do that with comments? It could easily classify comments that are part of a bandwagon and flag them automaticly hiding them or for human review.
We are able to moderate email but the concepts we use to do so are never applied to comments, I don't know why, this seems like a solved problem.