#Medical bias is much more dangerous than any damn weight

[zenaidamacrouras1]

My mother literally died because of weight based medical bias (doctors ignored deep vein thrombosis symptoms she went in for multiple times) (I know because I was in the room for two of them.) (I was 13) and....so yea fuck doctors.

Paper trail, paper trail, paper trail. If you have a hard time advocating for yourself, it may help to imagine you are doing it for the next person who comes along behind you. Maybe your doctor will end up being a little less fat-phobic. You might actually save a damn life.

(not that you owe these assholes anything so feel free to waltz the fuck out while dumping a container of gasoline and casually toss a match over your shoulder with a little smirk as you leave)

This is why fat shaming can have tragic consequences.

[itbeatsbookmarks]

(Via: Hacker News)

We live in the age of Big Data. Free-to-play games collect 300gb of data per day. Websites track every pixel you touch. There are so many A/B services that you can A/B test which A/B service is best.

There are three kinds of lies: lies, damned lies, and statistics.— Mark Twain

Bad actors can twist and manipulate numbers to say what they want. We all know how they play the game.

There’s another, more subtly dangerous side. The side where smart, educated, well reasoned individuals reach a conclusion that isn’t merely wrong, but is the complete opposite of right. It happens in frightfully easy ways.

Simpson's Paradox

In 1973 the University of California, Berkley was sued for bias against women applicants to their graduate programs. Men were favored to women by 44% to 35%.

Source: Wikipedia

The lawsuit triggered a study. The study results showed that not only were women not discriminated against, but that women had a statistically significant advantage! How is that possible? The data seemed clear. The answer is Simpson’s Paradox.

Trends which appear in groups of data may disappear or reverse when the groups are combined.

Here’s what happened. Some departments had high acceptance rates and some had low acceptance rates. Women applied to more competitive departments. Men applied to more accessible departments. Taken on the whole men had an advantage. When broken down per department it was women who were more favored.

Source: Wikipedia

This lawsuit actually happened and is one of the most famous examples of Simpson’s Paradox.

I find this paradox delightful. I like it because it doesn’t merely change or skew the result. It completely flips the conclusion. And it’s so easy to do on accident!

Kidney Stones

There are two different treatments for kidney stones. Which one is better?

Treatment A — 273 successful out of 350 (78%) Treatment B — 289 successful out of 350 (83%)

The correct answer is… Treatment A! Weird right? What’s the difference this time?

Source: Wikipedia

Kidney stones can be classified as either large or small. Large stones are harder to treat. Treatment A is better at small stones 93% to 87%. It’s also better at large stones 73% to 69%.

Each treatment was given 350 times. The critical difference is how the treatments were split small stones vs large stones.

Treatment A — 87 small / 263 large Treatment B — 270 small / 80 large

When taking a single average across all 350 patients Treatment A’s average will skew towards the lower success rate of large stones. Treatment B’s average will skew towards the higher success rates of small stones. Treatment A’s combined average is then lower despite being better at both small and large stones!

Onion Peeling

I view Simpsons’s Pardox like peeling off the layers of an onion. The top layer of kidney stone analysis said Treatment B was better. After peeling off a layer and considering small stones vs large stones separately Treatment A was better in both cases.

If we looked deeper again we might find Treatment B is better in at least some circumstances. Maybe it’s better for eldery patients. Or small stones in obese patients. Or large stones in patients with another condition. And so on and so forth.

Here is Simpson’s Paradox once again.

Trends which appear in groups of data may disappear or reverse when the groups are combined.

I love this because every layer you peel off can invert your conclusion. At first glance B is better. Look deeper and A is better. Look deeper still and B may be better again.

Video Games

Simpson’s Paradox applies to college admissions, medical procedures, and your video game all the same. Accurate data analysis is really damn hard!

Here’s a bit of a thought experiment. A common scenario and different underlying issues that I’ve all seen first hand.

Your players claim Sniper is overpowered. Because of course they do. But what does the data say?

Sniper averages more kills per game than other classes.

Well sure enough. Maybe your players are right. Let’s peel off a layer and see what’s next.

Sniper averages many kills at low skill levels.

Sniper is played less frequently at high skill levels.

Sniper is more dominant on certain maps.

We could start making adjustments off this data. There’s a couple of obvious knobs to tweak.

But we shouldn’t change anything just yet. We’ve not gone deep enough. We need to peel back more layers. Here’s seven different scenarios I’ve seen first hand that each warrant a different change.

Sniper is easy to play but has a low skill cap.

Sniper hard counters classes played more frequently by new players.

Sniper is too strong on certain maps due to long sight lines.

Sniper counters enemy classes naively played on certain maps.

Sniper is fine but a synergizing OP support class is played more frequently on certain maps.

Sniper is fine but skill rating system fails to promote high skill Snipers to high skill tiers.

Sniper is fine but the skill rating system incorrectly promotes mid skill Snipers to high skill tiers.

The last two points are my favorite. First, the lack of snipers at high skill play might not have anything to do with gameplay! Second, there are two relatively opposite conditions that lead to similar negative outcomes.

Theorem Theory

I’ve got this idea. An unproven theorem if you will.

For any given statistical result and conclusion there exists a data set that produces the same result but opposite conclusion.

I think anytime you’ve reached a statistical conclusion you need to ask yourself what if. What if you’re in the middle of Simpson’s Paradox? What if you peeled off another layer and that reversed your conclusion?

If you proactively ask yourself what if you may find your conclusion is correct. Or you may find you’re caught in a paradox and going the wrong way.

Conclusion

No matter how much data you have you still have to ask the right questions. It’s painfully easy to have good intentions but ask the wrong question and find the wrong answer.

Simpson’s Paradox is just one example how easy it is to get turned around. Being aware of it’s existence and constantly asking yourself what if is essential to staying on the right path.

Bonus!

Here’s one more example of Simpson’s Paradox in action. I felt like two classic examples with hard numbers were better for learning. But this is such a good story I had to squeeze it in.

This is all from a 2012 post titled Page Weight Matters. Chris Zacharias was a web developer on YouTube and took some time to optimize the video watch page. Over time it had grown to 1.2 megabytes making page load times unnecessarily slow.

With a few days work Chris shrunk the page size to a mere 98kb. He even decreased the request count and swapped out a bulky Flash player for a speedy HTML5 player. Everything was great so he pushed it live.

After a week of data collection the numbers came back and… the new code was slower! Page latency had increased. Despite being 10% as large it was somehow taking longer to load on average.

Enter Simpson’s Paradox.

To avoid a paradox it’s essential to know exactly what groups you are measuring. If you have a group that used old code and a new group using new code and you compare averages that doesn’t tell you a damn thing unless you understand the makeup of each group.

In Chris’s case the new improved code was getting a lot of new traffic from Southeast Asia, South America, and Africa. These places were averaging two minutes to load. Under the old code it would have taken them twenty minutes.

Chris’s code wasn’t just a success. It was a radical success. Twenty minutes was too long to be usable. Two minutes is slow but good enough. Entire populations of people who couldn’t use YouTube before were suddenly able to.

Yet the initial data analysis called it failure. It has to make you wonder: How many times have you been caught in a paradox and not known it? How many times have you used data to make the completely wrong choice? My money says more than zero.

[textualdeviance]

Re: that last reblog about false neutrality: YES. Balance is bias. Facts always lean toward one side or another, and any news org pretending otherwise is gaslighting you. 

My J school taught exactly that principle, thank goodness. Unfortunately, a lot of reporters--particularly in TV news--don’t have journo backgrounds that include this kind of training, nor have they had proper training in copy editing and fact checking. Way too many of them are just talking heads with a generic communications degree, if that. Used to be everyone in news had either a J degree or rigorous training, but once the big corps with their profit-driven agendas started swallowing up local radio and newspapers and running big cable news nets, journalistic integrity got thrown out the window in favor of the 1980s version of clickbait.

Because these supposed reporters are not fully educated, it’s assumed that they don’t have the skill and wisdom to make a judgment call about what facts are and who is most able to provide them. Thus the practice of getting quotes from all sides of a story, and then laying them out with (supposedly) equal weight, and letting the audience decide who’s right. Problem is: The audience mostly isn’t qualified to make those judgments, either, plus they look up to reporters as people who know more than they do, and thus they expect that the news will reliably tell them the truth. So if some jackass on Fox includes quotes from a Flat Earther in every story about NASA, they assume that reporter is telling them that the Flat Earth Society is every bit as qualified to tell the truth about Mars as an astrophysicist from the JPL. Adding confusion: The editorial and “debate” segments/shows that don’t frame themselves as different from straight news reporting. Used to be people knew that commentary and opinion pages weren’t the same thing as reported news. Now no-one has any damned idea what’s actual news and what’s just someone bloviating or a couple of people yelling at each other for the WWE version of reporting. After about 30 years of this, millions of people are no longer able to determine who’s a properly qualified expert and who’s completely full of shit, and an entire generation of news consumers has no fucking idea what’s real and what’s not.

Fast forward to the intarweebs age, and now news has been fully democratized. In many ways, this is a good thing. If everyone has access to a wide-distribution platform, it’s harder for gatekeepers with bad agendas to suppress a story that makes someone in power look bad. (This is part of why people who love propaganda want to kill net neutrality--if you can make it impossible for the plebes to load the pages with real news, it’s easier to control that flow.) Unfortunately, this also means that every dipshit with an axe to grind can call themselves a reporter and insist that their stories be taken just as seriously as ones from actual journalists. See Alex Jones. See Breitbart. See Young Turks. See U.S. Uncut. See the myriad sites run by homeopaths and other “natural” scammers passing off anti-science woo barf as legitimate information. Bald-faced lies are now being framed as fact, and far too many people have absolutely no clue they’re being lied to.

So how do we fix this? Well, it’s actually pretty easy:

1. Support your local newspapers and public radio.

As long as your local paper isn’t run by a massive conglomerate like NewsCorp or Gannett, chances are good it’s doing some decent reporting. If your local big metro paper is shit, look for ones from smaller cities nearby. Many of the weeklies are doing pretty good, too--even the ones that are part of the Village Voice parent company. Figure out who owns it, who the EiC is and what their background is, and then pay especially close attention to stories written by staff reporters (rather than wire services, freelancers, or stringers.)

Subscribe, if you can, or at least pay for a paper copy. If you prefer to get your news in digital form, turn off ad blockers when you go visit the paper’s site, so they can keep making enough money to pay their reporters and editors.

Any local radio that’s affiliated with NPR is probably a good bet, too, especially ones run by colleges. Donate to them if you can. Ignore virtually all talk radio. It’s an absolute cesspool these days.

2. Support the best of the national/international news orgs.

While they do have a slight liberal lean these days, the WaPo is one of the best national-news sources out there. I’d trust them over almost anyone else, including the NYT. For now, NPR is a close second, but whether that lasts depends on how much Trump fucks with it. For wire stories, take anything by the AP with a grain of salt, and pay closer attention to anything from Reuters, the BBC and Al-Jazeera. Many international papers also have good reporting. If you can read another language, look for stories from Der Spiegel, Le Monde, etc. If you’re looking at the U.K., be aware that they have some absolute shit there--ignore anything from the Sun, the Daily Mail or the Telegraph--but they have some good ones, too. The Guardian is particularly reliable. In Canada, the Toronto Star and Vancouver Sun are pretty good.

Some magazines are also good, and because of their longer lead times, you can often get far more in-depth reporting than the constant flow of glorified headlines you see elsewhere. Many of these have a strong East Coast flavor/bias, so keep that in mind, but for the most part, stuff from the Atlantic or the New Yorker is reliable. Ignore the big weeklies, though: Time, Newsweek, etc. They’re every bit as useless as anything else you’d find in a dentist’s waiting room.

3. Ditch ANYTHING that doesn’t do its own reporting, or doesn’t pay reporters.

News aggregators are the scourge of journalism. If the site you’re on is simply repackaging or doing commentary on stories that someone else reported, stop going there. This doesn’t include blogs or other places that are specifically designed for doing news commentary--and are upfront about that--more just the places that link to someone else’s story in the first graf, then have three more grafs paraphrasing or spinning what was in that story, and calling it reporting. That is not reporting. At all. If the person on the byline didn’t actually talk to any of the sources in the story, it’s not real news. It’s clickbait.

Likewise, some places may have a bit of original reporting, but because they don’t pay their freelancers, they should be ignored. HuffPo is particularly bad about this. They’ve even gone so far as to try to justify this by saying that paying their writers would introduce bias. HOLY CRAP NO.

4. Do your own leg work.

The ramp-up for this can be painful, but it pays off down the road. When trying to decide whether a given news org is worth your time, do some research on it. Find out who owns it, how long it’s been around, etc. Get some background on the EiC. Read some of its editorials to get a feel for where they lean. Look at some of its staff-written stories and see who they use for sources and how they frame quotes. See if they follow up any dodgy quotes with other sources refuting those. If a source seems questionable to you, go look them up, too. Could be that the head of Scientists for a Better World is actually some anti-vax crank who lost his medical license and is now operating a cult out of a strip mall. Some of the worst groups out there have names that sound legit--they do that on purpose to sow confusion. Make note of the icky ones, and avoid any news orgs that use them as sources. Also, see how often they run stories that read like slightly edited press releases. If they’re way too excited about some company or product or person, they may have literally just copypasted from docs they got sent by some PR hack. While press releases are useful for getting quotes or initial information, they have to be followed with real reporting.

Also: Don’t rely on your friends or family to give you reliable news (unless they happen to be journalists!) I’m sure Aunt Sadie is a wonderful person and means well, but if she insists that the article she read about how vaccines are dangerous is the gospel truth, chances are good you shouldn’t trust most of what she says about other news. There are a fuckton of well-meaning-but-misinformed people out there, and while they may be good sources for news about your cousin’s graduation, they shouldn’t be relied on to tell you a damned thing about what’s going on in Syria or whether the county water board has been taken over by corporate stooges.

(This caveat includes me, BTW. If all this seems like horseshit to you, feel free to look me up, too. I don’t expect my words to be taken on face value, and I’m happy to be transparent about my background and perspective.)

After a few weeks of doing this kind of investigative digging, you should be able to determine which of your potential news sources is going to be the most reliable, and you can then follow them on Twitter or FB or--gasp!--even buy their dead-tree editions if they have them, and rest assured that what gets in your face is going to be good information. Try to have at least two or three that you regularly follow. Getting a variety of angles is always a good thing, and some places are especially good for one subject or region, but not necessarily useful for other things.

The only way we get better, more reliable news is to pay for what’s already good, and stop giving money and clicks to the bad stuff. All news has to rely on revenue these days, so money alone doesn’t make a news source bad, but if you dry up the cash flow for the shitty stuff and start dumping it on the good stuff, we can eventually get news media back on track. To get good news, you have to be a good news consumer. Working for responsible journalism is a job for all of us.