The Perils of Constant Feedback | Hacker News Reader

The Perils of Constant Feedback

tjcx.me

79 pointstomjcleveland21 comments

The author makes a good point, but I don't think it's even the main part of the problem. In e.g. notifications cause people to open the app, they are optimizing the wrong objective function. Getting people to open the app isn't your objective. If you just want people to open your app as many times as possible, you should actually close the app so that they can open it again.

E.g. when an app sends me an unwanted notification, I open the app so that I can turn off notifications.

Similar to the p=0.05 issue in science, for every bad product decision someone wants to make, there is another bad objective function that it appears to optimize. Just because someone provides a justification doesn't mean it's justified. You need to make sure that changes are improvements in the things that actually matter.

That reminds me when I installed Tik Tok to see how it was. You could disable notifications for likes/comments on your videos, but not the trending videos notifications. I uninstalled it to get rid of the notifications

can someone uninstall Internet ?

Seeing that you let a tentacle of an oppressive surveillance-crazy human-slaughtering dictatorship infect your hardware, you got off easy. Are you sure the uninstall was complete?

> you should actually close the app so that they can open it again

Please don't give them any ideas.

They should optimize for money.

I'm reminded of a conversation I had with a friend about installing solar panels. When I asked how to aim the panels to optimize for the most electricity, he quickly pointed out it optimized for the wrong thing. "You point the panels for the power company, not for electricity".

And that's true. And the power company is doing the same thing, so peak power keeps shifting not for efficient electric use, but to get more revenue.

> They should optimize for money.

They already do. Which is what makes web browsing and web-enabled app usage the blissful delight that it is.

Actually the objective should be the greater good but nowadays who takes that into consideration?

What is greater good? Morality is murky territory. If we can make money as a leaky abstraction/proxy for the greater good, then that would be best. UBI is one way to do that - if everyone has some money, they have at least some power to influence producer behavior. Another would be to quantify negative externalities and tax them somehow. Similar to carbon tax. e.g, in this case, notifications could be charged.

On the subject of notifications, people are being bombarded these days. I've found that turning off most notifications helps me focus. Even Messages is turned off so that I only see them if look at the badge or go into the app.

For emails I've switched off the badge even, and I need to actively go in and look for emails in order to see if there's something new. I wholeheartedly recommend this approach.

It allows you to work or do something calmly for an hour or two without being interrupted.

Then it turns into everyone turning off all their notifications. What was originally a reasonable amount of notifications on a smartphone 5 or 7 years ago has now grown to such ridiculous amounts (often notifications for things I don't even care about) that the apps competing for my attention are cannibalize each other and consequently they've collectively destroyed the market for my attention, by me choosing to turn off all notifications.

Notifications have essentially become a "tragedy of the commons" problem. Everyone optimizing for their self interest ended up creating all losers.

Also the 2 factor and marketing emails. Despite reasonable efforts it's almost impossible to keep a non-high effort inbox zero at work and personal. At least authenticator apps are becoming a thing and adoption is rising.

But white papers post sign up and "marketing people" just flood the channel. I wish there were human only channels where any automation was banned. I guess like whatsapp.

Same concept as "can we get on a call to discuss my product". A) please dont cold call me, and if you do please just ask if now is a good time and B) as a personal rule I won't get on any call with a vendor who won't tell me how many digits his product sells for. In 2020 you can basically put your price on the website for most things or hint at a range. If you vary prices by so much at the same usage/feature level that you need to hide it from part of the customers.... theres something wrong with the product because really good vendors nowadays don't act like that.

I heard it's worse with younger peeple. The moment the doctors office gets a tablet they won't talk to anyone.

One more thing: yes I clicked on all 8 tabs of your webpage. Yes I knew you were watching on hot jar. Yes I got there white papers too even thought at best I would skim one to see how good the graphic design was. Please don't creepily make it so obvious you're tracking at the page view level. Not fun.

Except most people don't disable notifications.

It's very likely that even with a few contentious objectors, they still are net-positive-attention, even lacking mine and yours.

I have done the same for a few years now. No work on my phone either. I am not out of the loop, the loop is there whether I check for it now or at the end of the day, and got over the "Fear of Missing Out" pretty quickly.

One of my finest decisions.

I really enjoy this article, because as simple as the point is, there's something a bit novel taking what we've discovered in Machine Learning and applying it to general life outlook strategies. I've often had this intuition about the best strategy through life is having a dynamic shifting goal/feedback horizon between 'long shots' and 'short-term heads-down' life focus, but the article far better articulates the form of this strategy.

Another ML metaphor is this: don't just start from one point and do gradient descent. Try starting at many different points and working your way through the descent logic. It may be more trial and error and more dead ends, but you might indeed find your global minimum.

To speak like a human: try new things and don't give up on them immediately.

Another ML metaphor: I wonder if the concept of an identity crisis is a way to dislodge yourself from a local minimum? You know something better is out there, but you haven't done enough exploration so eventually you rebel and give up your comfortable little basin.

I love that identity crisis metaphor

I've been through it, and from what I understand that's actually what triggers it. If you don't try enough things in life, you can reach a state of crisis -- so never stop exploring.

Imagine an AI assistant that tried to maximize well-being across time, instead of engagement (aka arguing with strangers).

Imagine a job board that maximized its users income over time.

These would become essential services, right? Let's line up these incentives :).

Long delays make this difficult to attribute and also, difficult to correct if there's a design error.

The Bear Grylls analogy is hilarious! He'll do certain things wrong at a level you cannot even attempt.

Reply on news.ycombinator.com