Hacker News Reader

Comment by HumblyTossed | Hacker News Reader

> Zuckerberg vetoed a 2019 proposal that would have disabled Instagram’s so-called “beauty filters,” a technology that digitally alters a user’s on-screen appearance and allegedly harms teens’ mental health by promoting unrealistic body image expectations, according to the unredacted version of the complaint filed this week by Massachusetts officials.

> After sitting on the proposal for months, Zuckerberg wrote to his deputies in April 2020 asserting that there was “demand” for the filters and that he had seen “no data” suggesting the filters were harmful, according to the complaint.

So, why didn't instagram ceo provide that data? Seems a huge oversight for someone in that position.

View on HN

> So, why didn't instagram ceo provide that data? Seems a huge oversight for someone in that position.

Because there is no data that show causation?

Instagram does a ton of A/B testing on new features, I'm sure they could have done the research.

I’m sure they had A/B tested the feature, but it’s hard to think of a metric they could measure in those tests that would reliably detect the alleged harm.

I’m not sure they made the right choice in the absence of data, but I do believe that they didn’t have the data or a good way to get it.

Only way I could immediately think of is to survey users.

It reads as "survived users".

The idea that we can, as technologists, build or continue to do something that is so intuitively wrong because it is, in some sense, profitable and there is an absence of data to prove it’s harm is an indictment of modern society and ethics.

“Intuitively wrong” is subjective. The idea that a self-selected, self-righteous minority should set the moral compass for the majority is the real indictment of modern society and ethics. If there is no way to achieve majority consensus on one decision (there’s always a collective consensus on whether to be on a platform), then it only makes sense that the platform owners should be the one to choose

Yeah I think what you are identifying is also definitely wrong. But what I am saying holds.

You don’t think, for example, that most people would think it’s a bad idea to write software to facilitate showing teenagers unattainable and more desirable versions of them selves?

> You don’t think, for example, that most people would think it’s a bad idea to write software to facilitate showing teenagers unattainable and more desirable versions of them selves?

That ship sailed with the invention of the printing press. This conversation happened in the fashion industry with publications like Cosmopolitan long long ago.

The difference is in fashion magazine it is editors who choose what to project. With apps, it is teenagers themselves who create versions of themselves because kids compare themselves to others and it gets very clique - the apps aren't the problem per se, it is human behavior that is buggy and that is the real thing to fix because the next thing will come around that feeds off of the same suboptimal pathways in our brains, whether it is designed intentionally to make money or not.

People say this but what, you're going to A/B test a feature removal? Randos being like "hey, where is this feature now?"

At least if you remove the thing wholesale you can do the PR push along with it, and actually do something _good for people_ (at least if you subscribe to this concept in general).

I think it's hard to suss out some stuff quantitatively on this, but ... honestly? If you do qualitative interviews with a bunch of kids across the spectrum, have quotes with them, an info packet with all these articles.... maybe you're wrong, but the upside feels important and the downside feels pretty trivial.

I remember when Facebook did research to show causation between newsfeed items and mental health. Everyone reviled them for “running dangerous manipulative experiments on human beings”.

After that, I bet they definitely want to avoid the same bad PR. It would be worse if they ran it on children, and someone died during the course of their experiment.

The problem wasn't necessarily the A/B testing rather than what came next, Cambridge Analytica. This current headline seems to follow the same message, that being able to manipulate a users mental health is beneficial to Meta.

Collecting the data isn't a problem, but utilization is? Seems to be the same problem.

Facebook did A/B testing on psychological effects of their products a decade ago and people are still up in arms about it: https://www.nytimes.com/2014/06/30/technology/facebook-tinke...

Doubt we’ll ever see them do it again.

Because if Mark doesn't want to see data, then it doesn't exist.

Yeah, he never saw the Rohingyas get slaughtered.

"If we have data, let’s look at data. If all we have are opinions, let’s go with mine."

-Jim Barksdale

You’re going to use company resources to create a test for your position or pull the data to support it (using resources that could be pushing forward on the business) when the CEO hates your position? That’s not generally a wise strategy for corporate executives.

Even if you have the data, this is likely all about the CEO not really wanting to hear the argument or go in this direction and not his officer’s inability to provide supporting data.

I'm seeing a lot of responses about the morality or PR implications of trying to A/B test this, but this seems fundamentally impossible to A/B test to me and points at a bigger problem with what companies and their marketing departments believe they can know and the limits of what science can actually do.

The hypothesis here is that usage of professional but automated editing tools to make people look more beautiful than they really are promotes unrealistic competition standards and makes people feel worse about how they look in a regular mirror that is showing them the truth.

How do you A/B test this? Giving the feature to some people and not others isn't good enough. The impact is because of seeing other people look more beautiful on the Internet than they actually look in person. How are you going to prevent a user from seeing the photos of other users who use filters? Even if the global social graph had strict partitions, which I imagine isn't the case, these photos leak into web search, news articles, and the same technology gets adopted by other platforms. You can't A/B test something that impacts the entire broad culture. The world is too connected.

> “All the people that I’ve talked to internally about this were like… Mark’s level of proof, in order to be able to take the work seriously and act on it, is too high,” Bejar added. “I think it’s an impossible standard to meet.”

Reply on news.ycombinator.com