NotAtWork

The methodology you linked to might as well be called "how to lie with numbers".

The methodology's way to assign numeric values to things doesn't relate to intrinsic things about the objects under study, but rather, is a numeric score reflecting the subjective opinion of the reviewer.

That something uses numbers doesn't make it objective, and this is a prime example.

This article seems to measure something orthogonal to learning styles, namely that multiple styles combined and variety/novelty also play a key role.

Can someone explain to me what the data in this article has to do with learning styles as they're commonly thought of?

As a tutor, I've found that certain styles of explanations click more intuitively with some students than with others.

Do you have a mechanism to explain this?

Thank you.

That step had me really confused.

I would trust lots of professionals will my bitcoins more than I'd trust myself, especially if they had the right insurance policies in place.

The problem is that Coinbase only insured themselves against damage from coins-in-transition being accidentally lost or being stolen. That doesn't actually cover most of the cases that I'm worried about, because while a 3% loss would be annoying, it's nothing compared to the other failure modes and should be a loss that's sustainable by Coinbase independent of this insurance policy.

The problem isn't fundamentally exchanges holding the coins, but rather, exchanges holding coins so irresponsibly currently.

The bitcoin insurance policy I want from an exchange is where they insure me against loss due to actions taken by the exchange and underwritten by a major underwriter, and not the exchange itself. Only then would I really believe that the exchange meant to protect my bitcoins fully, because they've essentially put themselves on the hook for the damages (by having to bear the cost of the insurance policy, recouped by some kind of holding fee or transaction fee).

The exchange having insurance does relatively little to protect me against loss, even if it's a smart business move on their part to limit their losses to creditors in the event of a theft of inventory. That's only sane business, and it's a little crazy how long it took that to become normal in bitcoin.

It seems (naively) reasonable that the exchange could float something like 10k-20k policies of some mixture of 0.1btc, 0.5btc, and 1btc protection, depending on the level of activity in an account, and require that anyone holding more bitcoin than that in their trading account take out a separate policy for that amount or else acknowledge that they're only protected up to 1btc. (An additional corporate policy against theft of inventory seems reasonable, and likely would lower the rate on these separate policies, since the underwriter would know you could cover some degree of loss already, through that policy.)

Edit:

Thinking about the numbers, it would only be ~$20 million of coverage to cover 50,000 customers (40,000 at 0.5btc and 10,000 at 1btc). This seems reasonable.

> The insurance covers losses due to breaches in physical or cyber security, accidental loss, and employee theft.

I'm not convinced that this even protects the online wallet against Mt Gox style corporate malfeasance, where much of the damage was done as an intentional corporate decision by the leadership. (I'm conjecturing that's what happened at Mt Gox; regardless, this seems not to cover against them trying something intentionally with their coins related to trading, and then losing them all.)

Additionally, as you pointed out, this only covers a very small portion of their holdings - a percentage smaller than the number of Mt Gox coins that were "recovered" (about 25%).

Finally, Coinbase being insured against the value of the bitcoin loss doesn't translate in to you being covered, as Coinbase may accrue other debts they have to pay off before your holding in the event of a hack large enough to cause a serious business disturbance, such as paying off creditors with higher standing.

From their website, they appear to both provide services to talk to inmates (of which I have a few friends from my time growing up in a poor, crime infested neighborhood) and to act as a data broker about the prison system and inmates, something I've had a long interest in and is related (tangentially) to my professional work in data analytics.

I won't be using their services either, neither as a communication method nor as a data supplier, and will try to steer any companies I work on data analysis at away from using their data.

Looking at the data, I don't think the author did anything with it than see initial numbers matched her feelings, and then called on people to undertake a massive, actual study because she just know this is it.

I certainly think that there are problems with gender in society in virtually every place we could examine, and that we have a long way to go before things are what anyone could call ideal.

I just have trouble with a lot of the statistics used in these discussions, and find that they're very often 20+ years out of date (ie, from or before 1994-1995), don't control for confounding influences, make misleading comparisons, etc.

I would take posts like this much more seriously if she posted the dataset, but I'm not sure how she could do this without revealing personal details or editing the text (which likely would bias the choice of recipients further, or could introduce a new bias). I would even settle for the details of how she did the bucketing, correlations between words and numbers of entries per person, etc.

The short answer to why I think that this article isn't a real source of data is that the study in it has about the statistical power of just asking everyone who's a friend of a friend on Facebook for people with a moderate number of friends.

Everyone already knows that there's a problem with gender in tech. This article does nothing about saying where it is and doesn't really contribute anything to the topic.

Yes.

You have a sampling bias in the types of people you associate. Facebook applies a filter to posts which selects for things that are popular with a wide swath of your friends and which support your political/social views (as guessed by their profiling tools).

That something blows up with your friends on Facebook is usually a better indicator that it's polarizing drivel than that it's a well thought out, impactful study, since that's what the machines (essentially) optimize for.

I don't actually want to be arguing against this study, but without knowing how those two fields are quantified, those numbers are completely meaningless and referring to fuzzy numbers like that is one of the hallmark ways to lie with statistics.

There are also lots of uncontrolled variables, such as the average woman submitting 1.4 reviews, while the average guy submitted only 1.3 reviews. This means the totals would obviously be off, even if everything were symmetric per capita, a fact not mentioned when the numbers are displayed.

Further, it's likely that the discrepancy in the number of reviews per capita submitted is a sign of some underlying sampling bias, which needs to be accounted for before we can really talk about the distribution of feedback.

I think this is a serious issue that needs addressing, but that's exactly why I feel it's important to object to bad math.

Most of the comments (all of them that I've seen) say that the effect is real, but the source isn't necessarily a bias in the reviews, because of how the study was conducted (eg, there could be a hiring bias).

Bringing up a different cause to explain an effect is not dismissing the effect as existing in the data, merely calling in to question the source of it.

This article lacked the power to tell if the effect (ie, different language in reviews) was correlated to their gender or because of a selection bias in the people studied, eg because of a possible difference in the distribution of personality types in the field from each gender.

It's possible that there's a hiring bias, and reviews are being done fairly, but show biased results when naively studied because of the underlying bias.

This is what makes social statistics hard, and there are many faults in studies involving gender that fail to account for possible confounding effects in the data.

Why do you think the likelihood is very small?

Has there been research?

I see a difference between SpaceX and Tesla in this regard: Tesla had a production car catch on fire with a customer using it, while SpaceX broke their rocket doing an experiment with rockets during development.

It seems pretty reasonable you break your things while you're testing them out - I mean, we've all see those car crash commercials, right?

Most of my development time is spent on correctness, safety, and refactorings.

If the extra information helps those at the cost of not one of those, it's almost always the right choice for me to use.

I can see several ways that you could implement storage and processing using that, thank you.

For anyone else thinking about the problem, there are two processes that could be used by item sorting to do processing:

1. You can use the stream of items through a pipeline (if deterministic enough) to encode a bit of strings. You can do this without enchanting, because you only need two item types.

2. You initialize an instruction decoder and have it sort the item based on type, which then triggers a command block to decode the instruction further.

I've worked in hardware, and think this kind of play is meaningful.

You confront many of the engineering challenges you'd face (in terms of line routing, timing challenges, design concerns, modularity, etc), but they have simpler and more straightforward constraints because of the different medium.

You're also really unlikely to get results using exactly this same skill set when trying to work with real hardware. The immediate need for programming and complicated tools becomes apparent when you try to grapple with the complexity of modern ICs.

If you're just going to have them blink LEDs using logic gates, I fail to see what you think you're getting from hiding all the logic circuitry in little black boxes and making the parts harder to work with. I've encountered all the engineering concepts (minus some physics concerns) I learned in several quarters of circuit design as an undergrad.

The reality is that Minecraft actually is a very good tool for introducing hardware concepts, because it removes the dependence on understanding software to do something.

(Really, the tooling around hardware needs to be better, particularly for things like FPGAs. Raspberry Pis and Arduinos are making good inroads.)

How do you read off the state of the NBT data?

Can command blocks match items on NBT data or chests on their data?

> You do, as it happens. In order to operate, all machines must pass fairness and randomness testing. They are also under a spot-check regime to ensure there are no post-manufacture changes or defects.

Many of the machines actually aren't random. Rather, large batches of results are (semi)randomly chosen such that there's a particular distribution of results in the next several thousand (or a million) results.

That page you linked is horrendous!

25 million Americans live in the 10 biggest cities.

33 million Americans live in the 20 biggest cities.

44 million Americans live in the 50 biggest cities.

I think any of those is a reasonable number to build a rail system for.

> did they act in your interest?

To address this, we have to reintroduce the complexity that I had omitted in my comment: there are other people besides me and the (figurative) person I was addressing.

I think the honest answer here is that yes, the CIA staff did what people wanted, likely in violation of what the law actually was, because they saw it as their duty to undertake such tasks when the public clamored that they needed safety from a dangerous world, everything else be damned.

One of the hardest and most interesting questions about government is how do we stop ourselves from making collective bad decisions on effectively a whim in a moment of weakness/fear/etc?

> Let different US states have different laws (that the Federal law cannot trump) and let's see through experimentation which set of laws is best for different sets of people.

This only makes the chunking smaller, not fundamentally removes the chunking that you objected to. ("Chunking" in this sense meaning that we have a region ("chunk") that has to all follow the same law picked by some fixed (possibly local) strategy, such as majority wins or other common voting strategy.)

I actually agree with the concept that the chunks should be smaller about many things than they currently are in the US, and that the US generally has too many federal laws.

The problem with your answer to the question posed, however, is that we still aren't sure what we mean by "best", except just to ask every body and take some kind of "most popular answer wins" or averaging strategy. Fixing a heuristic like "best" allows you to deduce (most of) the rest of the things you need to figure out.

One obvious potential problem with this strategy, which is the point of federalism to address, is that the equilibrium reached by allowing each chunk to vary its laws and people to move freely between them, is that we might settle on to a suboptimal equilibrium (or never converge) which is considerably worse than one we might have discovered via planning.

Obviously, the current state of affairs isn't what many people would think of as optimal either. So the question, at least as I see it, becomes "What mechanic do we focus our effort on to fix?"

As someone who thinks that voting is a useful mechanic, my answer would be that addressing the voting mechanics would allow the law to more accurately reflect the desires of people governed and allow for better selection of heuristics for decision making.

Your stance seems to favor just removing large segments of what exists, and seeing what gets invented to fill in the gaps.

If one person in every family made their in-laws grumpy about the CIA torture scandal, then the entire nation would (at least once, over the course of months/a year) have an actual discussion about the situation. That discussion then goes on to inform their views on various topics, including who they should vote for in the next election.

Government officials, at the end of the day, are chosen because the circles get filled in on ballots, and the circles get filled in from a choice made on the aggregate of our experiences, including discussions with friends and family.

The difference is that if we talk about it, even if it makes people uncomfortable, then the nation discusses these issues and takes them in to account when voting. If we don't, then the nation doesn't.

The rate at which pissed off people bother their family members about what they're pissed off about directly influences the course of national politics, even if the particular result of any particular discussion is some pissed off in-laws and a small percentage chance a bubble gets filled out differently in a few months.

> Why can't we have a system where we argue things out and have some heuristics for finding out what the best decision is for a given group of people?

We do this.

We just haven't resolved two fundamental problems:

1. The heuristics are not faithfully stated nor evaluated. This actually has upsides and downsides, as it allows them to be unexpectedly flexible and comprehensive at the cost of some corruption and unpredictability. This may just be a fundamental trade-off of such metrics. The heuristics are just currently embodied in little sacks of meat with some extra processes happening, because of stand-alone computers being a relatively new thing. (I'm actually not sure how much better with regards to lack of corruption and transparency we could make the heuristics than "just pick a guy to choose", because of limits in decidability and other related computer science/math.)

2. We don't know any way to define 'best outcome' besides ask everyone what they think, and resolve conflicts with some kind of strategy like "go with what the most people think is the right answer".

So I really don't understand your position: it's essentially that you don't see value in participating in which heuristics are chosen, and think that people should pay more attention to your wants and desires, even if you don't advocate for them with the same strength as other people.

Honestly, your position mostly sounds poorly thought out and entitled.

> > If you don't like the options, it's worth asking what you, as a citizen, have done to improve them.

> I can revert that and ask you, what have you done, as a citizen, that you've made the choices so awful for everyone else?

I picked people who would act in my interest, and since you didn't participate, they only care about my interests, not yours. Why should I care about your interests, when you can't be bothered to represent them as part of the process?

Edit: Your comments seem to not really reflect the scale and complexity with which the system you're talking about operates.

I mean, if you care more about not arguing with your relatives than you do that the CIA abducted and tortured people, then you have little incentive to engage with the wider political process.

Of course, most people think that the US government abducting foreign nationals and torturing them at black sites is a considerably bigger problem than their mother-in-law being grumpy at them for a couple days, and want to do something about it.

I can't tell if you honestly are asking for answers, because all of those have fairly direct answers from people who saying voting matters.

Also, the one about war is a strawman, because it conflates war and the draft, which are related but not identical.

Ed: Typo.

> Where is the line between optimizing an ad pipeline and violating privacy?

Right at the start.

Of course, the allegation was that Facebook broke privacy laws, and not just that they violated people's privacy, so your response is a total non sequitur: you're talking about how it's good business to violate privacy, which doesn't address at all whether it's good business to break the law, nor whether these laws draw a meaningful boundary.

Unless you meant that anything which is in the business' interest to do shouldn't be illegal for it...?