Hacker News Reader

Hacker News

Comment by mhh__ | Hacker News Reader

Parent Full thread

mhh__

> Continuous feedback of mood

That's very black mirror.

Also, didn't facebook get busted basically manipulating feeds to see how they affect people emotionally?

View on HN

Busted has a few meanings. They were caught, yes, but faced no apparent penalty, either legal or user backlash.

Yeah these ideas always sound great in theory, but the privacy issues are immense. Apple however does seem like a company fully committed to privacy. Of course this is much easier for them as their business model is selling hardware and doesn't rely on selling your data to advertisers at all.

> Apple however does seem like a company fully committed to privacy.

Trust but verify

I never understood why those in the tech industry were upset about that.

They were doing A/B tests and were tracking sentiment. So what? How is that different than running an A/B test and tracking any other dimension?

People were asking questions like "how many people died because facebook was manipulating their emotions?". What? Why can't we say that about any change ever? It seemed to only matter to people when they started looking at sentiment.

I agree, but probably not for the reason you think. All A/B tests on users are a form of psychological experiment and should require explicit consent, like an ethical psychological experiment would. It’s concerning that the software industry gets away with this unethical practice until the experiment explicitly mentions mood or sentiment, instead of implicitly being used to change user behaviour.

That's at least a principled stance, which I can appreciate.

I still disagree, though. It seems like the only logical way to get out of your thought process would be to create software and then never update or change anything. Either that or have every user sign a waiver saying that the product could change in the future and that might impact them psychologically.

After all, why is it different when you're making a change without A/B testing it? Surely every change has some hypothesis attached to it. Does it only matter to you when developers observe the metrics changing?

> I still disagree, though. It seems like the only logical way to get out of your thought process would be to create software and then never update or change anything. Either that or have every user sign a waiver saying that the product could change in the future and that might impact them psychologically.

This is a false choice. There are multiple ways you can ethically approach improving software, such as:

- user interviews

- A/B testing with informed consent

- user testing (directly observing the user while they use the software)

- focus groups

- feedback forms

- crash/error logging

There are probably more, but the theme among all of them is user consent. The ethical implications of using engagement metrics to drive software development is well-worn territory at this point, especially with respect to social media, and I don’t understand why the software industry seems to get a pass when others don’t.

you could make a study with willing partecipants

Think about it, they did experiments knowing negative sentiments like anger, fear, anxiety, sadness, despair were likely potential outcomes. They were so prepared for those negative outcomes to occur that they went out of their way to measure it. They did these experiments against thousands (millions?) of users.

Those outcomes are likely on any A/B test. Just because you don't measure it, doesn't mean it doesn't happen.

Except they induced the experiment. If your A/B test might negatively affect millions of people, and the only upside is your quarterly bonus, maybe it is immoral to be running those tests.

is any AB test ethical if done with unwilling partecipants?

what if any A/B testing is actually similarly as bad? Users didn't sign up to be your test subject...

Maybe we should do studies where subjects are consenting to it...

Is that different to you than, for example, releasing a change to a product to everyone and observing the change in metrics?

if that change is the same to every user, then I would start my argument from the position that it is, in fact, different;

if the user can tell that something changes because of a new features/changelog that explains it (or lists it for more trivial changes), again, seems different.

I would be open to discussion on how it's the same, or how it's unethical, though

Reply on news.ycombinator.com