KidSwitch: Identifying Kids in Apps and Sites | Hacker News Reader

KidSwitch: Identifying Kids in Apps and Sites

blog.superawesome.com

42 pointstdaltonc34 comments

I assume the implicit goal here isn't really "disabling features" like private messaging, but rather replacing age-gating in apps like Steam that stand between a non-logged-in user and the landing page for a rated-M game. (Or, maybe for the first time ever, actually effectively preventing a five-year-old from accessing a porn site just because the porn site doesn't want to be party to showing a five-year-old porn, rather than because the parents installed a content blocker.)

But—possibly-dystopian question here—could this be used in the opposite direction? To keep adults out of safe-space-for-children communities (like the now-defunct Club Penguin)? "Prove you're not an adult" instead of "prove you're not a robot"?

I can't imagine a usecase for such a feature. It would prevent parents from auditing the page.

It would be just as easy to allow adults only read-only access, without allowing them the ability to comment.

This would solve e.g. the Elsagate thing. "YouTube Kids" video? Adults can't be creepy in the comments, because adults can't even post in the comments.

I think there are two use cases: the first is a naive "to prevent pedophiles from having access to children". If you don't think too critically about it, I can understand the enthusiasm for it. The second, and I think potentially more legitimate, is to offer children a space to socialize with their peers without having to worry about parental investigation. Whether or not you agree with the parenting style, it's an underserved market segment. A side benefit could be having your juvenile social media sealed somewhat like juvenile criminal records. Something like offering you the option when you turn 18 to delete your profile, or archive it with read-only access to those over 18.

Legally, I think it would face troubles in some countries, but countries with laws granting more self-determination to minors might make it a useful environment for exploring ideas with your peers without constant fear that someone may now, or in the future, be offended by your teenage-angst fueled rants. I know I would have enjoyed that, I had a constant fear of retribution through highschool that prevented me from significantly engaging in social media. It wasn't particularly out of line, the usual "my parents suck, my school sucks, let me be a poorly adjusted adult right now!" kind of thing.

But I also grew up during the start of the "social media is forever" age, where it started to be common for years-old social media posts by celebrities to become common. And the posts are always judged by today; not the person you were then, or the environment you were in.

Frankly I still maintain very little social media presence. I would love to interact with people like that, but I also know that all it takes is one slip-up or vaguely phrased statement to sink you anymore. I like that this might be am opportunity to let kids be kids, and say the darndest thing; but still allow them to be full-fledged members of an online and in-person community without fear that their half-baked childhood ideas might ruin their future.

I think the purpose is advertising targeting kids disguised as child safety.

> KidSwitch works by evaluating a number of signals in a content session (gesture patterns, content interaction, hardware inputs etc).

These are all the sorts of things that advertising platforms already do.

> Prove you're not an adult"

And I think any machine learning based on that is a false sense of security.

Kids are humans and there's nothing inherently unique about them that distinguishes them reliably from adults enough to not have a by-pass feature. I would bet an adult who wants to "fool" the system would be capable of doing so.

> But—possibly-dystopian question here—could this be used in the opposite direction?

This already exists, I think. I mean, how many adults hang in the chans? How many even know which are the good ones? And then there's the whole coded-slang aspect.

Sadly enough, adults you find there are mainly perverts, cops, and extremists trolling for fresh meat.

More than you'd think

Accuracy stats? This seems like an anti-feature to me, and a little paradoxical to collect tons of data to exclude others from data selection. I'm sure it could work, but it seems like there could easily be a couple years standard deviation which would lead to many false positives.

It will probably be amusing to see the amount of decidedly non–kid friendly abuse hurled at this by adults who happen to like, say, watching toys being opened.

The most popular video on LinusTechTips is the unboxing of an RC fire truck. I think people do enjoy some toys being unboxed.

That sounds less like adult subscribers to the channel being really into that video, and more like tons of non-subscribers (i.e. kids) finding it by searching for toy unboxings and it happening to rank highly because of the relative popularity of the channel. So, not really an example of what "people" do.

> (gesture patterns, content interaction, hardware inputs etc)

It doesn't sound like it takes into account which pieces of content you watch, just how you navigate the app. So I don't think watching toys being opened will cause you to be detected as a kid.

So children who want to use social features will have to browse more adult-like content to adjust the age estimate? I'm not so sure this will have the intended effect

Nothing will stop driven teenagers. Nothing.

The same way people of my generation had CD/DVD ROMs full of pirate games, software and pornography - teens today will find the path of least resistance.

>Historically the approach to this issue has been age-gating (asking a child to confirm they’re over 13), which makes everyone feel better but doesn’t actually solve the underlying problem. > >Today, we’re unveiling KidSwitch, a new piece of…

Which underlying problem?

I’m more interested in education & intelligence levels, iff* we’re to regulate content over web.

E.G. you must be thiiisss smart &or demonstrate n% content comprehension to continue, return, reply etc.

*iff as ‘if & only if’

I think the underlying problem is that parents want the entire internet to babysit their kids for them.

edit: Maybe what we need is a kids.txt to go along with robots.txt. Some file on websites with some kind list of what interactions a site has (can you leave a message, chat, access hardcore pornography, etc) so auto-blockers can block (or even just limit usage).

So what if a site doesn't have a kids.txt? Block it I guess? Prompt to go into some special mode (no video, no posts, alert parents)?

I guess sites lying might be an issue, then you'd either need whitelists or blacklists. Maybe tie compliance to whether the browser bar goes green.

This used to exist in browsers in the 90s. Site owners never bothered implementing it. https://www.w3.org/PICS/

Scanning it, it seems PICS specifies an rdf format that ratings agencies can use to define their own rating system, with self reporting (of various PICS implementations?) being an option for sites. I can imagine some site owners would find it a little hard to implement.

It wasn't that owners never bothered implementing it, rather people were actively against it because they felt that no content should be filtered period, and never mind children.

It was actively campaigned against to the point that if you did implement it you were criticized in public. So sites stopped using it and browsers stopped checking it.

Some linkes: https://archive.nytimes.com/www.nytimes.com/library/cyber/di...

https://www.nytimes.com/1998/01/19/business/technology-to-le...

https://www.wired.com/1997/02/microsoft-employs-good-clean-p...

> I think the underlying problem is that parents want the entire internet to babysit their kids for them.

It's virtually impossible for parents to check on exactly what their kids are doing, all the privacy controls that the tech world is so proud of have made it almost impossible for parents to keep an eye on their kids.

So what do you expect from parents? What else are they supposed to do? It wasn't even 15 years ago that the whole world helped shield children, now suddenly, nothing.

Not just nothing, active work to prevent parents from watching their kids, coupled with sites also refusing to filter anything.

And then parents get criticized for not doing enough........

There are regulations in the US against accounts and data collection on children under 13.

I wonder if KidSwitch violates this. I mean, it actually claims to track user behavior. Isn't that "data collection"?

I once signed up for a chess website that asked a non-trivial question to enable signup. I didn't know the answer, but with a quick google I was able to find it.

I think that's great. One must either be passingly-familiar with the subject matter to participate, or be reasonably intelligent enough to google things you don't know. Some mild gate-keeping does good for a community.

The underlying legal problem is the fact that in many jurisdictions it is forbidden to advertise to children of a certain age, sometimes with restriction on the time of day and / or the products advertised.

Toy companies and other producers of children aimed products do what they can to stay on the right side of the legal line while at the same time maximizing exposure.

Here, have an amusing 4chan thread about this

https://warosu.org/sci/thread/8913970

There should also be a /v/ version of this but with filtering bosses/sections.

It would be nice to see some technical documentation on exactly what's being looked at, but I doubt it's available given that it would just allow users to game the system.

Beyond privacy implications, what are they doing to make sure those with physical differences such as disability aren't negatively effected?

This isn't the kind of technology that focusses on actually getting it right, this is a product meant to be sold to companies that need a way to off-load their legal responsibility of preventing children below a certain age from using certain features.

This product ticks a box, it doesn't actually prevent children from using the features it acts as a barrier for. It probably works quite well in prepared demonstrations and with content that fits neatly in certain age groups. Those demonstrations won't include people with disabilities or quite mature children of course.

Speaking of legal responsibility, there is also a legal responsibility to not discriminate (even if inadvertantly) against the disabled. This seems like a pot of gold style quick fix, which is why I have strong reservations about its viability.

This is creepy as hell, since you need a lot of surveillance on children to train this thing.

How could this avoid discrimination against atypical neurologies?

I'm not exactly sure what "atypical neurologies" includes.

But I suspect that some of my personas would rate age~13.

Why use Machine Learning to identify kids? Simply ask "Identify the Karate Kid actor"

if they pick Jackie Chan over Pat Morita, you know they are a kid 100%.

The use-cases for this are obviously pretty wide. The really galling thing about this is that YouTube(Google(Alphabet)) didn't start offering this 3 years ago.

Reply on news.ycombinator.com