Hacker News Reader

Comment by techwizrd | Hacker News Reader

I can see why this would be useful, but we already spend a lot of time chasing down data quality issues in AI/ML models and statistics as-is. We may be requesting permissions to workaround/correct GPS accuracy issues, comply with local regulations, or turn on a feature flag. For example, if you're on your home network I should hit HomeAssistant via a different route than if you're away from home. Or maybe local regulations require me to limit certain features, deny access, or confirm age. Or it may be necessary for fraud prevention, such as determining whether your account is being accessed from an unusual device or location.

I could see this leading to more data quality issues, legal liabilities, and difficult bugs. We should be able to deny access or provide coarser information (e.g., an age range rather than an age or a county/state rather than fine-grained GPS), and applications should be designed to handle those gracefully.

View on HN

I understand all of those concerns, but... as a user they're not my problem.

For example I don't care about where I am and neither should you for legal issues or features. This results is situations like me going no holiday and the bank app failing to work/install because they don't have presence in X country. Wanna know if the usage complies with regulations or I want a feature? Ask me, don't guess.

Wanna know how to access my HomeAssistant? Ask me, because there's a ZeroTier network making HA available directly wherever I am.

For fraud prevention it's not necessary to know more of my tracking info. It just makes it easier for the company at the cost of privacy issues for me.

I get the developer's view here, it's harder when someone feeds you wrong data. Tough, we're supposed to deal with that as devs and maybe just ask the question directly rather than guessing through other means. You should be doing that already, because GPS is not always reliable, contacts change, people have complex lives in terms of which laws apply to them.

As a user, I don't care about your ML model. I care about not sending you personal info. Coarser info isn't good enough - you can convince me you actually deserve my real info, or you can get no info (preferable), or you can get fake info (alternative if you degrade or break my user experience because I wouldn't give you that info).

I understand there are some legit use cases for validated info. Unfortunately, targeted advertising and other types of profiling are also common use cases for the same info. It's a lot like MAC randomization on public wifi - it sucks, it breaks legit use cases, but it was needed because too many companies were using it to track people.

You should care. My point is give me coarser info or no info, but not fake information—and I say this as a researcher working on synthetic data generation methods to improve privacy. Fake information is a pain for developers _and_ users. Unreliable fake data has real, significant (and potentially disastrous) consequences that are often difficult to anticipate.

It sounds like the assertion here is that the data is used for reasonable purposes and improves the user experience. That's likely very true.

If I prioritize privacy over that better user experience, though, that should absolutely be something I'm allowed to do.

I might not want to give a stranger at a bar my real phone number, so I give them a fake one. Maybe this leads to a worse experience than if I just said "no thanks", but I really value the fact that I'm allowed to do that.

Reply on news.ycombinator.com