Comment by DanBC | Hacker News Reader

Comment by DanBC | Hacker News Reader

Parent Full thread

Comments are often toxic, and there doesn't seem to be a way to fix that.

It's impossible to know without watching something whether it's full of swearing or homophobic / misogynistic / transphobic hate speech.

Ads are not targeted to the actual viewer, but to some combination of regular browser user and main youtube user. this means that children will be shown ads for alcohol or gambling. (note this goes against what Youtube says here about complying with local law: https://support.google.com/adwordspolicy/answer/6012382?hl=e... Note that this is something all ad networks get wrong, and it's likely they're going to see regulation in the uk at some point if they don't fix it.)

It's hard to know before clicking a link whether it's sponsored content or not. (in the UK sponsored videos need to be clearly identifiable before the user clicks the link or clicks play. in the US sponsored content needs to be clearly identified).

If two people (eg a parent and their child) watch Youtube the recommendations fail hard. If a single person watches youtube the recommendations are pretty poor.

The way to fix comments is clear: human moderators. Since this can't br automated (yet), Google doesn't do it.

It's not clear. Human moderators are not the magic pill that fixes everything. Google has taken enough flak for injecting far too much liberal bias into their system and human moderators will only outrage more people. You may see it as righteous when opinions you don't like are swept off of websites you use. Everything is neat, tidy, and non-challenging to you. It's a bit Orwellian to me.

Google has an automated API in beta to detect toxic comments. I've been playing around with it and it works pretty well and could solve the human moderation problem.

I don't think human moderators are the clear fix at all.

Both Facebook and Twitter employ a large number of human moderators, and they still regularly make judgment calls that are comically bad - defending stalkers while suspending their victims, suspending users that are the target of coordinated harassment campaigns, etc.

The reality is that when you have an army of people who need to churn through a vast amount of content per day, without the time or authority to look at the full scope and context of user interactions, they will make lots of bad calls. Doubly so if the moderators live half a world a way and have little awareness of the culture and norms of users they're asked to evaluate.

Add this to the fact that the really obvious abusive comments can and are already being moderated by automated systems, it means the human moderators are overwhelmingly exposed to the trickiest, least cut-and-dry situations.

> defending stalkers while suspending their victims

or refusing to delete photos of women and their ex-partners when they've been killed by their ex-partners.

http://www.bbc.co.uk/news/uk-england-gloucestershire-3461822...

Is there any form of communication where you know whether is full of things you disagree with before receiving it?

Yes.

eg http://www.bbc.co.uk/iplayer/cbeebies

Reply on news.ycombinator.com