Hacker News Reader

Comment by scrollaway | Hacker News Reader

Being mad about something like this is just choosing the wrong hill to die on, honestly. Studies (https://support.mozilla.org/en-US/kb/shield) are clearly just a/b testing of features for the browser. Doing this opt-in creates selection bias which corrupts the whole idea of running a/b tests.

So instead of saying "you should run your a/b tests opt-in only", what you really mean is "in order to fight for privacy, you should never run a/b tests", which, again, totally wrong hill to die on. Why is it so important they not be allowed to try different features to figure out which ones work and which ones don't?

Mozilla has made so many missteps they circled a small planet with them. This isn't one of them.

View on HN

If you advertise a browser as being all about privacy, you absolutely should not be a/b testing users without their explicit consent. It is amazing to me that this needs to be said.

It needs to be said because some folks are blinded by their belief in the virtuousness of themselves or their in-group.

I may agree with the goals, but that doesn't mean I should blindly trust them to execute them, right?

Okay, why? What does a/b testing have to do with privacy?

If I told you right now that HN is A/B testing the size of the vote arrows, and you see larger ones than I do, how exactly are your privacy rights violated?

A/B testing implies extra tracking and reporting of behavior, to distinguish between the A group and the B group's response to the change.

On a website, that isn't too much of a problem, because you expect your activity to be tracked since you are interacting with a web server.

In a browser, the same tracking is more intrusive. My behavior on a site is now being sent to the browser author, an unrelated party.

All A/B testing really implies to me is, at most, crash reporting, which shouldn't be much of a privacy concern.

I've never seen A/B testing used for crash reporting. Perhaps it could be used for this, but it's flabbergasting to me that you see it as only for crash reporting.

You've never seen partial rollouts of a feature?

> My behavior on a site is now being sent to the browser author, an unrelated party.

But you're just assuming it is. It could also be "On uninstall and crash, report back which studies were enabled" to track how many people are uninstalling their browser when the flag stupid_feature="on" is set.

Be mad about the data collected [specifically, what type of data is collected], not about the concept of a/b testing. Set clear boundaries of what is and isn't acceptable. Otherwise you're fighting an ever-uphiller battle. Because if they want to know which sites you visit and the privacy boundary is "a/b testing isn't acceptable", they'll do it via history syncing instead.

> Be mad about the data collected, not about the concept of a/b testing.

What does this sentence mean? A/B testing fundamentally requires data collection, so I don't understand how to be mad about the data collected but not A/B testing.

I think they're saying to be mad about what data is collected. For example, be mad if they're storing your IP address and location and not be mad if they're just storing a counter of how many people visited their marketing page.

A privacy first browser should be completely up front about what data is being collected. The simple fact that they did while implying something else means that they arent to be trusted to not abuse the privilege in the future.

That makes sense.

Don’t be mad _that_ data is collected, be mad about _what specific_ data is collected.

I don’t know that I agree, but it does help me understand the meaning of the previous post. Thanks!

> But you're just assuming it is

Of course we are. We didn't consent to being experimented upon or to have our behavior analyzed, so we have no basis for trust.

> Set clear boundaries of what is and isn't acceptable

Acceptable depends on trust. Ask for permission, not forgiveness, and I am more inclined to trust. Without that, I assume the worst.

You're not assuming the worst if your assumption is that "some third-party I distrust will respect the checkbox they themselves control".

Mozilla has arbitrary code execution rights on your computer the moment you allow them auto-updates. They're also entirely free to ignore your opt-out or opt-in requests.

That's not a analogous situation. You can inspect and control what the code does on your device, but you can't inspect what is done with your data after it leaves it.

Yeah... no. If your claim is that the code on your own device can just be inspected, then you can inspect what data leaves your device instead of assuming Mozilla hoovers up everything including your porn stash, credit cards and the weed hidden in your cd tray.

Yeah... yeah. What data goes out and what is done with it after it leaves are absolutely not the same thing.

> My behavior on a site is now being sent to the browser author, an unrelated party.

What is the consequence of this? In the abstract I understand that it doesn’t feel great, but I’m asking in the specific, how does it make a difference to any individual user?

If you knew that the user in the A group made fewer misclicks after the back button was shrunk … how does that harm anyone?

You could say the same thing about pretty much all adtech tracking, since nearly everyone would be unable demonstrate any concrete harm to themselves.

That's not the point. The point is unless you are doing data collection on anything but an opt-in basis, you are not pro-privacy. What you do with the data or why you collect it or if that collection is harmful is irrelevant to this principle.

That's just extra data points that you give away for no good reason. Realistically it will not amount to anything (like, I'm sure, absolute majority of data collected even by sleaziest adtech). But worst case scenario? You data gets sold to a medical insurer, deanonymised, fed into ML models, then used to justify increasing premiums based on inferred health info.

A browser A/B testing and sending analytics data back home is different from a website.

A _privacy marketed_ browser sending such analytics data back home is different from a regular browser doing so.

Obviously you'd need to know how they're measuring the effect. And I think that's the problem: Firefox doesn't make that easy enough. If Firefox gave on installation

* a short list of all the data they could collect, and

* a promise that they won't collect anything else without asking for the user's consent again,

then I doubt anyone would complain.

Informed consent is a standard practice for human experimentation. It's a serious ethical lapse that the web companies have decided it doesn't apply to them simply because it's inconvenient. Here's some references:

https://researchbasics.education.uconn.edu/ethics-and-inform... https://www.interaction-design.org/literature/article/conduc...

Research ethics is way too deontologically focused, it needs reform. See: at minimum tens of thousands of deaths directly caused by failure to use challenge trials.

This is often taken too far. If the research is "try out an experimental medicine" then yes, of course you need informed consent, because the patient is taking a risk.

If it's "see which color blue people like best in your UI" then this is nonsense. Nobody is going to be harmed regardless of which color blue you pick, so they don't need to be informed. Some privacy zealots may want to be informed, but they don't have a reasonable argument for why they need to know.

It doesn't need to be so life or death here. This is a privacy focused browser with the explicit message that it cuts out tracking, so the tracking without consent is pure hypocrisy.

It absolutely should apply to psychological experiments as well. If you can't see that treating fellow humans as lab rats for you to experiment on is deeply problematic, then you're an actual sociopath.

I don't care how you rationalize it, as long as there is any sort of intervention, you need informed consent.

If changing a color in a UI is a "psychological experiment" then all artists and UI designers are doing psychological experiments all the time. Are you doing a psychological experiment every time you edit a CSS file?

This is redefining something ordinary to sound bad, but that doesn't make it bad.

Secretly monitoring unaware people to see how they behave based on the UI design is absolutely experimenting on people, and deeply problematic. It doesn't matter if other people are doing it. It's wrong. People aren't for experimenting on.

It's experimenting on software, not experimenting on people. "People" aren't the part of the equation that anyone cares about in this scenario.

And I do not think you're even sort of having this argument in good faith. You can't possibly be a rational person and believe that changing UI colors is "deeply problematic experimentation on people".

Forgive me if I'm a little pissed off but this privacy extremism shit is exactly what leads to people looking down at anyone who is a privacy advocate as a nutjob. So, thanks a lot for that, more work for those of us out there who want, to, I dunno, protect citizens from being genuinely fucking spied on and their data passed around like candy so better ads can be served.

But hey, it's all good, you're keeping those pesky UI designers from knowing which of their redesign is better, so at least we got that going for us.

It seems a bit dismissive to wave off pp's discomfort with becoming an unwitting UI test subject.

I also would not agree with any argument like "opt-in is necessary to reign in anti-privacy activities of bad companies like Facebook, but we're good UI designers at Mozilla so we don't need to use opt-in and can use opt-out with impunity."

A browser that collects and reports any information on its users without their explicit permission is simply not putting privacy first.

Any privacy-focused browser should make UI testing - or any other telemetry or analytics - opt-in and should generally clearly disclose any information being collected (and preferably make it available for inspection) and also describe all parties that will have access to the information as well as what it will be used for.

The fundamental problem is your view of human beings as tools to be used to further your goals. Other people aren't tools for you to use. They are your peers.

If you view them as instruments, you're a villain. It doesn't matter if your intentions are good. Everyone has good intentions. Every atrocity in human history has been made with good intentions.

It doesn't matter if you don't think it will do them any harm. That's not for you to decide.

except for that one time they silently pushed an extension marketing a TV show, which is probably how anyone who's heard of Firefox studies found out about them.

https://www.theregister.com/2017/12/18/mozilla_mr_robot_fire...

Yes, I'm aware. This is one of their million missteps (and this was a particularly bad one). But that their marketing people are absolute idiots has no bearing on whether a/b testing as a concept infringes on privacy. They could have used any other thing, they're literally in control of what runs on your computer the moment you allow them auto-update rights.

Marketing people are always absolute idiots. The failure here was allowing them to push that extension, and that wasn't their choice.

Why do marketing people have control over the a/b test?

Reply on news.ycombinator.com