admiralteal @ admiralteal @kbin.social

Posts

0

Comments

795

Joined

2 yr. ago

With group policies, it means that the insurance companies can do their actuarial work on the entire group in aggregate without having to have considerations about prorating based on certain individuals entering or leaving the policy throughout the year.

At least ostensibly. I doubt this actually happens. It's mostly just a way to limit administrative overhead for both the insurance companies and the employers.

Don't think for a moment that employers don't like the whole open enrollment system too. Even if they CLAIM it is a PITA, it lets them only have to deal with this work for newly-qualified employees, separations, and otherwise only once a year.

Either way, it's of no benefit only harmful to the actual consumers of the insurance. But since individuals aren't the customers, that doesn't matter.

It's the persistent lie of the free market.

Yes, in theory if health insurance were a transparent, understandable product that you could easily switch with another one as an individual based on your needs and costs, market competition would optimize that product rapidly for service and cost.

But every single thing about that theory is wrong.

1. It's an intensely opaque process. You have no real way to know what the costs are going to be, what your needs will be, what your options will be. You can't even know what a doctor's appointment will cost under healthcare until you get the bill, sometimes a full year later (at least in my experience. Nothing about healthcare costs are understandable even to someone with an advanced medical degree. The layperson has no hope.
2. You also cannot easily switch it with another product. Open enrollment and contracts severely limit you. There's only fixed, stressful windows where you can change it -- and even then, you're back to point 1. What is the difference between the two plans in actual practice? It's all just gambling.
3. As you already observed, if your employer offers healthcare, you basically have no choice but to use that product because the subsidies are so intense. You are not an individual. The individual plans suck, are intensely expensive, and usually both across-the-board. The ONLY affordable option for the average person is the employer-offered product, so your choices are severely limited.

And it go this way because the most powerful agents in this system are not individuals.

With home/auto insurance, basically everything gets driven to a commodity product because all costs and risks are pretty uniform and predictable. That's why there is vanishingly little difference in the core products being offered by these kinds of insurance companies, and why the idea that switching your plan is sure to save you gobs of money is... improbable, outside of just periodic renegotiation of rates.

With health, the costs and risks are WILDLY unpredictable. The difference between an "expensive" customer and a cheap one is many, many orders of magnitude.

So naturally, risk must get hedged. The system's need for efficiency is going to try and package people together, just like any other high-risk, low-reward financial product. The need to group people is obvious, so we made it mandatory for employers to provide insurance as a weird workaround to the logical thing of government-run insurance. Now the customers are primarily employers who have TOTALLY different needs and desires from cheap, high-quality healthcare service. The free market will now do its work and optimize based on supply and demand. The efficiency gains will benefit the vendors (insurance companies) and the customers (employers). Individuals are not benefiting from market forces at all.

Free markets are great systems where they apply. They're really good for rapidly assembling efficient systems to get products to customers. But they only work where they work. The persistent lie of the free market is that EVERY problem can be solved with a free market. Nope. Only certain problems can. And where free markets don't work, that typically means you have need of government to step in instead.

Nor will Hamas release all of theirs, frankly.

I had some French cousins we would talk to a little bit at the time, and I remember their descriptions of the early internet were just absolutely bizarre in comparison with the minitels.

In those days, I'm sure every major region and country had vastly different experiences.

But yeah, at least my experience in the US was that AIM was huge. My entire peer group was connected through AIM. That and memorized land line phone numbers.

If you modify a stick into a shiv and stab someone, you won’t be arrested because you “modified an object that you own” – you’ll be arrested because your modified object was then used in a crime.

Not to stretch the metaphor too taught, but in this case the guy going to jail was the guy who runs the social media for a business that sharpen sticks for folks that don't know how to do it themselves, not the guy actually doing any stabbings.

God, I remember in the early-ish days explaining what browsers were to AOL users.

It honestly felt pretty early in AOL days that people were mostly just using it for email, chatrooms, and otherwise as a web browser on the regular, non-AOL internet. Then AIM becoming more popular as time went on, but eventually third-party clients totally obviated that in a lesson Google would learn from well (and their takeaway was to destroy Jabber/XMPP with great prejudice before they lose control over their users).

Explaining parents that all they needed to do was open another browser -- literally any other browser -- while AOL was running and they could go to the websites with it was rough. "AOL has you connected to the internet already, you don't need to use it to go to infoseek.com" or whatever.

Whenever they finally did it it seemed like magic. WOW, how does this connect to AOL! Then they'd close AOL and disconnect the modem and tell me the other browser was broken.

I remember all my friends convincing me to switch to Opera because it had tabs and that was revolutionary.

Do you understand that there is a difference between an optimal outcome and an optimal algorithm?

Please don't project this onto me. The difference between these two things has been my core argument with you. The algorithm can be optimized to get a certain result. Your claim that the algorithm can ALSO define the desired result is preposterous circular reasoning. It's clearly not even something you believe, yet you continue to argue with me about it.

Whether the result is good is the entire point of the field of numerical analysis

Gibberish when you cannot objectively define what a good result is. Numerical analysis has many purposes. In this case, the purpose is to efficiently achieve a result based on input criteria. Whether or not that input criteria defines is "a fair election map" is a subjective question that reasonable people can and will disagree on. That disagreement is going to be political.

You can quit the condescension. It doesn't impress me. I've been involved in data science research on similar optimization problems, so I know full well how it works and what the shortcomings are. The big one being that you need to know what is being optimized for. And "the most fair election map" is not a quantifiable outcome. You remind me of the PI on my last research project, handing me a giant dataset and saying "What does it mean?!?!". Just couldn't understand that I can't give him an answer unless he has a meaningful question, and just couldn't understand that p-hacking to come up with something publishable was pointless and dishonest.

You have your hammer and you're SURE this problem is a nail. You just don't even know where the jobsite is.

You only get to say what is desired in the outputs.

Yep, correct. So if you have a partisan process for deciding what is desired in the outputs, how are you going to get a good result? Which is why my entire point from the beginning was that chasing down magical algorithm solutions is pointless if you do not START with addressing the political problem.

In this case, Wisconson's process for redistricting is partisan. That's why this thread exists. That's the problem being discussed. That's the original sin. You're not going to datascience a solution to that problem.

And moreover, these huge redistricting controversies all have that same thing in common -- a partisan distracting process. Like magic, the issues seem to largely go away in places that don't have a partisan redistricting process. GEE, WHO WOULDA THUNK IT.

Why wouldn’t you assume a panel of nonpartisan humans for both?

It's not an assumption. That is our starting point here. You need to listen to the people you are talking to and understand what is being discussed.

IDGAF if you want to use machine learning to design maps. That's fine. Could be very helpful in making the maps, I'm sure.

But only AFTER you have limited the effects of politics in the decisions about the algorithm design. Otherwise, all that machine learning is still going to get you corrupt maps. They'll just be really, really good and and hard-to-understand corrupt maps. Corrupt maps that achieve the goals and obfuscate the means they became corrupt through layer after layer of mostly-opaque and essentially impossible-to-comprehend models, graphs, and networks.

I'll end this discussion by summarizing it for readers:

Me: if a partisan process is used to build maps, algorithms won't get you fair maps. Start by addressing the partisan process.
You: algorithms can be optimized to get you perfect election maps and they can know what election map is perfect because they optimized to get a perfect map. This other guy is a dumb idiot who doesn't know how algorithms work.

Depending on the particular cannon, in some of them the supervision is active at all times. That his senses just let him see all the layers and through all the layers.

So it's not just that he CAN see all of them all the time. He doesn't even have a choice. It would mean nudity to any degree was extremely unremarkable to the man.

The way optimization works is that you have a desired outcome, and you adjust inputs until you reach an optimal outcome.

Maybe I am confused, did someone else say this?

Your entire belief here is predicated on the belief that there is an objective, optimal outcome. And there isn't. If there were, you don't need to run the elections in the first place.

You're confusing two different things. You understand, correctly, that a well-designed bit of machine learning can get you efficiently and reliably to a certain result. What you don't understand is that whether that result is good/optimal or not is a subjective question the computer cannot answer. The computer can only get better and better at tuning its various coefficients and neurons in order to get it closer to the result it has been TOLD is optimal.

The only decision that really matters is the decision about what optimal is. Which is a political question, in the case of redistricting.

You’ve already listed several desired outcomes yourself...

Correct. There are lots of outcomes I think of as desirable. Some of those outcomes are in direct opposition to each other, too. Conditions where increasing the voice of some group reduces the voice of another. Which ones I care about most is a question of my personal politics.

Over and over again I repeat, the decisions about how to weigh these metrics against each other is a political one. One where political actors allowed to take on the problem in a political process will chose the inputs that get them to the results they want.

If you can flip a seat by saying that having proportional racial representation is twice as important as having proportional religious representation, a political actor that wants to flip that seat will tell you that proportional racial representation IS twice as important as proportional religious representation. Either way, you have either a race or religion NOT getting its full proportional representation reflected in the outcome.

And it's not even just the proportional outcome that matters. The decision about how competitive to make the races is ALSO an input here with no objectively correct decision.

she has accused the Satanic Temple of making filings that “are only meant to evoke strong emotions and incite others.”

I don't think anyone in the Satanic Temple would disagree with that. That's honestly a pretty good description.

It's basically a rite of the religion, to stir up trouble. To rile people up until they reveal their hypocrisy.

You run an optimization algorithm to figure out how to identify an optimal algorithm? That's begging the question. It's circular reasoning.

You said it yourself, you need to have a "desired outcome" in order to optimize. And the decision about the desired outcome is a political one, end of story. A political agent will optimize for the political result they want. Using an algorithm is of no advantage if you don't change the external factors first.

There is still no objective measure of what is optimal for this problem. But I can tell you aren't listening.

that is an optimization that can be specified without human intervention

It is absolutely not. That is not true. The optimization is a political question because what is "optimal" is a political question.

If you have a theoretically-perfect algorithm, you don't even need to have the elections. That's an unnecessary abstraction; it can just pick the winners directly.

Leave the personal attacks off. There's no technophobia here -- I just have a different (and I believe much better) understanding of the technology's inputs and outputs.

With a nonpartisan or multipartisan approach, there is no need for a "best" group of humans. You have innate checks and balances on the political process. Without a nonpartisan or multipartisan process, you plainly don't.

Close to $2B, as I recall, but they're already sold to the Saudis.

Good job or bad, Trump will find a way to not pay you for doing it. Of this, there is no more golden rule.

These ideas all sound good to me.

But also, these system choices are a political ones that affect political results, and so whether that effect is positive or negative is subjective to politics. There's winners and losers being picked in every system adjustment.

I don't particularly disagree with what you're saying. My criticism is that building these elaborately-complex algorithms that are deciding how to build the districts... it's having the political apparatus chose the voters instead of the voters choosing the political apparatus.

In simple terms: a data-science approach to redistricting is TOO powerful. It is too effective at getting to a result. It fundamentally undermines the democratic process.

In a well-functioning democracy -- which I am not going to pretend the US is -- elections need to be competitive. They need to allow for surprises. Political arbitrage. Some amount of unfairness is the price we have to pay for that. And a theoretical algorithm that sufficiently takes into consideration EVERYTHING we need to ensure flawlessly "fair" elections where outcomes have a perfect proportional match to voter opinions/identities (and how do we even know they truly do in the first place?) may as well be trusted to pick the winners while it's at it. We're just in Foundation at that point.

No, I'm not. Again, the video I referenced if you want to know what most influenced my thinking on the subject.

The problem is, the algorithms only factor into consideration the things they're programmed to consider. They have to also be told how to weight and consider these properties.

By relying on algorithmic approaches, all you have done is change gerrymandering from the art of line-drawing into the art of algorithm design. It's not really any different in terms of outputs. All a bad actor needs to do it figure out the initial weights to specify -- which can easily be made to look innocuous -- to get the election results they want. And with any remotely competent approach, they can work backwards from their result to discover initial weights that will achieve that goal. So long as the process is allowed to be partisan, this will be what will happen. Right now they're fucking overt about it because they have been able to get away with it. In the near future, they'll be WAY more cagey and subtle about it and lawsuits like this Wisconsin one will become harder and harder to win.

And merely deciding what weights you do and don't consider and how they compare to each other is an inherently partisan process. There's SO MANY criteria for deciding which groups should not be split or can be split by these algorithm lines. This is just cracking and packing in a different form.

On the flip-side, a transparent and non/multi-partisan process resists these kinds of machinations in the first place. A more manual line-drawing process is much harder to do partisan cartography stenography with.

The data scientists should stick to evaluating the maps. They should stay WELL clear of designing them.

Someone with Reagan's exact politics in this day and age would be considered a left-wing woke monster by most of the modern right.

Dude did a lot of tax hikes and was against having capital gains taxes be separate from income taxes.

He did try to genocide the gays though and modern conservatives do like to genocide the gays.

Is it a "good" outcome if you have the right ratio of dems to republicans, but zero black representatives when black people are a quarter of the population? If one in seven statewide voters is hindu, surely you should have 1:7 hindu representatives -- but if the hindus are spread out that is sure to not happen unless districts are DESIGNED to select for one of those reps, which is going to water down everyone else's votes to a meaningful degree even without changing the proportional representation.

Or how about maps that are designed to get you just the right representative breakdowns by whatever standards you want, but are designed to create nothing but no-contest races where the incumbent party has no chance of losing? That's a pretty bad outcome that is honestly SUPER common in redistricting.

There should definitely be a high correlation between seats and votes, of this there is no argument. But there's a lot of tension in trying to be fair about maps. Too much to pretend there's an algorithm that can nail it. The gold standard is to have a complex, transparent, and non/multipartisan process to make those compromises