Hacker News Reader

Comment by mochomocha | Hacker News Reader

I'll offer a contrarian opinion. A lot of the issues raised recently in our industry around privacy or algorithmic bias come from the fact that as an industry tech has the hubris to believe that it knows better than its users what is best for them. Under the pretense of simplicity and scalability, we think that every individual behavior and tastes can be reduced into a one-size-fits-all product and business model, sprinkled with machine learning to give the illusion of a product experience tailored to different people, when in fact it is only mildly so. And I'm saying this as someone building machine learning models for a living.

View on HN

I don't see that as a contrarian opinion, at least not with my reading of the post.

Bicking isn't saying you should decide exactly what your users want and give them that and only that. (Well, he's not saying to not do that either, if in a situation where that's doable...) He is simply saying that if your final conclusion is "users want more control", then you're copping out. You are giving up on figuring out what your users really want (whether they know what it is or not), and just throwing options at them in hopes that they'll be able to make something out of them. Or at least stop complaining.

Giving your users control isn't bad. Not figuring out why your users are asking for more control is bad.

Most users don't want control - more things to think about takes energy and time. The users that do want control, however, are likely to be the users championing your product. Take a standard transmission in automobiles - the extra control it gives isn't necessary, and a large number of consumers prefer the automatic transmission. But the racecar drivers and car lovers are much more likely to want a standard. If you have intelligent, invested users, they'll want control options, and they'll usually tell you what they should be.

On the flip side my government insists on giving me a choice of schools for my child. I don't really want choice, I want the local school to be good and for my children to go there.

In this view, users still don't want control, its just a symptom of the failure to provide what users do actually want.

The truth is probably between these 2 outlooks

> I want the local school to be good and for my children to go there

That's assuming there's a universal "good" - even if a school meets some general standards one school can have a better arts program, the other might be more biased towards sports, etc. I don't see anything wrong with having choice - having a good baseline everywhere is a separate issue.

Well lets go out on a limb and say all schools are good, and specialise in one subject. Which should I send my 5 or 11 year old to?

I mean they like football, but they also like drawing, but then neither of those are well known for paying the bills. So I still circle back making sure they get a balanced education.

The problem though is that choice is the supposed solution to not having good schools. That may be UK specific though.

The idea (although rather poorly implemented in the UK state system) is that choice leads to competition, and competition leads to improvement even of those at the bottom (and even for those who don’t really pay attention to the choice/don’t care). It’s a market philosophy of sorts.

I know that's the idea. The result is schools spending a non trivial amount of time jumping through hoops for the benefit of people who are ill equipped to evaluate those schools. On top of that the number of places are fairly static year to year so all schools end up getting filled anyway, but the 'better' schools are filled with the students that arguably least need the benefits, whereas the worst schools are filled with the children of the least able.

Yes. The 'market' mechanism for popular schools to find money and permission to expand while less popular ones are closed or taken over is missing and subject to too much political interference. So it arguably ends up as the worst of both worlds.

Ah, exactly the issue that sprung to my mind too. The tyranny of "choice" as introduced into various UK governmental systems is so bloody annoying. Great, I can "choose" my kids school, armed with a spreadsheet and looking at various progress, attainment and wellbeing metrics, that never actually seem to capture how happy and fullfilled the kids are at school. I can "choose" my hopsital consultant using the choose-and-book system.

No. I want to send my kid to the local school that will be good and I want to go to the local hospital where I will be treated by a well qualified caring medical professional. I don't want to have to create pivot tables.

I'm British too, is this a British thing? Schools, not pivot tables.

Knock yourself out https://www.compare-school-performance.service.gov.uk

It also provides the strange incentive to optimize for glossy powerpoint presentation that communicates our core values to parents. Organizations which never had to have a marketing department finds themselves struggling without one.

>On the flip side my government insists on giving me a choice of schools for my child. I don't really want choice, I want the local school to be good and for my children to go there.

Yeah, people mostly start asking for control when you make the default experience shitty. They want control to escape the hell they've been put in, but they have to pay a price for it either in money or time. So functionally, "give users more control" is really saying "make the user pay more to use my product."

In an ideal world everything would be bespoke and custom tailored for me, but the idea there is that the tailor knows how to create what I want. Not that I have to go tailoring all my own stuff. I'm not competent enough to do that unless it's my hobby.

> They want control to escape the hell they've been put in

OTOH, if "user control" is seen as table stakes or baseline, it can limit the shittiness level that users will tolerate by lowering switching costs.

In Germany consumers certainly do not prefer automatic transmissions (I do and it puzzles me why most people don't). So there's at least some culture/history involved in these preferences. I'd assume "wanting control" is also a bit of a cultural issue.

In the end, I think/fear that on average lazyness wins every time.

automatic acceleration feels really neutered in every car i've come across, being in control feels better than waiting for the car to kick in automatically.

then again i ride bikes too and prefer a more active driving experience

I do think there is more of a feeling of attachment to the machinery with manual transmissions.

Germany is a huge exporter of precision-engineering so there might be a cultural sentiment there.

Manual transmissions are the default in most of Europe. They're cheaper, they're more efficient and they make the most of small engines with limited torque.

>They're cheaper

Not substantially.

>they're more efficient

Not any more.

>they make the most of small engines with limited torque.

Ding. Ding. Ding. We have a winner!

People like manuals because they can choose when to up-shift whereas the guys programming the automatic transmission will program it to up-shift as soon as they think they can get away with (for fuel economy and emissions). With a manual you can let your 1L hornets nest scream along at 4k all the time giving you close to peak power on tap whenever you want it. Basically manuals are a hack around regulations that force manufactures to build cars with performance characteristics nobody wants (under-powered engines and transmissions that up-shift at the drop of a hat). Those regulations are much stronger in Europe so the value proposition of a manual transmission is better there.

force manufactures to build cars with performance characteristics nobody wants

Well... performance characteristics that everybody wants -- for everybody else. They themselves want warp drive, but prefer everybody else help hold down the price of gasoline and the carbon content of the atmosphere.

Most of the time I drive economically and up-shift very early. Sometimes I want to have some fun and fly out of corners at 4 or 5K on my small 1200cc car.

Can't do that with most automatics.

I wish they were cheaper. I paid a premium for an automatic. It also limited the vehicles I could choose from.

> I do and it puzzles me why most people don't

The feeling of slamming into second as you come out of a corner in a little sports car. In an automatic, its just a pedal and the wheel, in a manual you physically engage the warp mode. It feels good, gives a greater sense of power in your actions and also contributes to enjoyable familiar muscle memory. Humans dig physical ritual.

That is a separate issue from control - one of defaults. You can have a simple default that works or a complex one with defined behavior which have usability issues.

Customization gives control to those who want to take it.

Automatic transmission cars are very unpopular across Europe. In all seriousness, I don't know if I've ever even seen an automatic in real life, except for travelling to the US.

There's a big difference between "choice" and "control". iOS doesn't allow you to customise your home screen, but it does offer strong guarantees about what happens to your personal data. My local convenience store has a hundred different kinds of sugar-sweetened beverages, but no fresh fruit. My cellphone company offers dozens of different price plans, but they're so confusing that I strongly suspect that I'm being ripped off.

Choice only amounts to control if those choices are meaningful and legible.

Perhaps I am thinking too narrowly, but I think the algorithm-driven approach is much less hubris than someone at the company making a decision. These models reflect what users do -- isn't that much more democratic than employees deciding?

There is an orthogonal argument that this is also harmful to some people's aspect, but "give the people what they want" doesn't seem like it's reductive. Especially when it seems like people in our society can be clustered and have some clear groupings occur.

Is the algorithm-driven approach really any better, though? If our UI leads users to accidentally take an action, our data will actively mislead us into thinking it's what they want!

Designing clear interfaces, measuring things and coming up with models are all really hard. Data needs to be paired with human judgment and decision making.

>These models reflect what users do -- isn't that much more democratic than employees deciding?

What users do is different from what users would like to do. If people's aspirations lined up with their real behaviors, Planet Fitness wouldn't have a business model. But when navigating the world people often want to improve themselves into something they want to be, not fall deeper into the worst tendencies they exhibit in any given moment.

You aren't wrong, but that's presumptive on humans being able to create algorithms that aren't implicitly bias. 2, hopefully not overly simplistic examples, A) Automatic Paper Towel dispensers that don't work nearly as well with people with dark skin. (Unintentional bias on the hands of the designers) B) People getting google banned and then being totally helpless and useless and not being able to get themselves unbanned (an algorithm determined there was abuse, and there is no recourse to tell the machine that it was mistaken, the entire thing was automated away via algorithms.)

This is absolutely not true. This is what they ACT like, but this is not their true motivation. These companies want all control to be theirs, and they want to be free to do what they want with users and their data, and they want to spend as little time, money and effort as possible while doing it. They know they can't state this honestly, so they dress it up in excuses, rationalisations and outright lies.

They KNOW they don't know better than its users. But they don't want to let users decide, because users would make decisions that go against their own profits, so they remove the choice and lie about why.

A lot of the issues raised recently in our industry around privacy or algorithmic bias come from the fact that as an industry tech has the hubris to believe that it knows better than its users what is best for them.

A good way to summarize, "users want control," is to note that, "users don't want to be screwed." Users don't want to be put into a situation where they have lost so much power, they don't have practical options, or the only options they have left are punitively unpleasant.

>A lot of the issues raised recently in our industry around privacy or algorithmic bias come from the fact that as an industry tech has the hubris to believe that it knows better than its users what is best for them.

I don't think this is that bad, on its own. Choice is a pain in the ass, so it's useful to have things pick for you, when you don't care to choose yourself (which is very common).

The main issue is that there is never offered a means to correct the algorithm, when it chooses something the user decidedly doesn't want. That is, there's no user-inputted feedback mechanism; they go for implicit tracking like "eyeball-time" to correct the algorithm, despite the user (me) being both able and willing to give an explicit answer. I want Netflix to give me better recommendations, and I'm willing to put in a bit of effort to support that.

But Netflix doesn't want me to help them help me.

And thats just stupid.

Netflix used to have those explicit feedback mechanisms via user star ratings, and even ran a contest to improve their recommendations based on the data[0], but it turns out "that explicit star ratings were less relevant than other signals. Users would rate documentaries with 5 stars, and silly movies with just 3 stars, but still watch silly movies more often than those high-rated documentaries."[1]

Changing from stars to thumbs increased the quantity of data (at the expense of granularity), but didn't change the relative advantage of tracking behavior over explicit ratings.

Now, your ratings may be free of these issues, but that data isn't very useful in isolation, so blame the bad data from other users for the removal of the feedback mechanism.

[0] https://www.netflixprize.com/

[1] https://variety.com/2017/digital/news/netflix-thumbs-vs-star...

So two competing priorities in a) giving users all the control/configurability they want and b) per OWASP and most SQL security standards “never trust user input”

Explain straddling that line ... oh and the third axis is cost.

Those two are perfectly compatible actually - although doing both is harder and thus economically more expensive.

Not trusting user input is about security. Verify that everything is something which they /should/ be able to do. Think say for a game's bank account. They should never be able to directly change that value without valid transactions to and from.

If they can lose money by writing a check to null that is bad. If they can /gain/ money by writing a negative check to null that is /really/ bad. Also really bad is being allowed to just take money from banks of others without a validated transaction.

Configurability lets them do anything that is in their rights. If they can access their bank balance and set it as part of their signature that is allowed along with say setting their crosshair to display the target HP which they already know. If they can create a custom texture of their bank balance repeated ad nauseam that is fine. So long as they are capable of that stuff already.

You can actually get emergent conflicts from this in somewhat unexpected ways. When users are enabled to export a bunch of private data in the name of choice and control, they can be encouraged to do so unwisely. Users rarely stop and read warning messages, no matter how simple and clear and straightforward they are.

You see it as hubris. I see it as earnestly trying to build amazing things but failing.

I agree with you here. The majority of the people involved are trying to do something good. I’d wager the egomaniacs are a much smaller percentage than those who want to build better things.

Are they not functionally equivalent?

I suppose morally there's a difference, but that's just us and our biases.

Hubris is a wonderful adjective for the arrogance now main stream in the industry.

Reply on news.ycombinator.com