Profile cover photo
Profile photo
Olli Orajärvi
Olli's posts

Post has shared content
Get it done

It's better to do something imperfect that helps than not help at all.  We so easily forget that.  Here's a great story to help us remember: the Hair Dryer Incident, as told by psychatrist Scott Alexander:

The Hair Dryer Incident was probably the biggest dispute I’ve seen in the mental hospital where I work. Most of the time all the psychiatrists get along and have pretty much the same opinion about important things, but people were at each other’s throats about the Hair Dryer Incident.

Basically, this one obsessive compulsive woman would drive to work every morning and worry she had left the hair dryer on and it was going to burn down her house. So she’d drive back home to check that the hair dryer was off, then drive back to work, then worry that maybe she hadn’t really checked well enough, then drive back, and so on ten or twenty times a day.

It’s a pretty typical case of obsessive-compulsive disorder, but it was really interfering with her life. She worked some high-powered job – I think a lawyer – and she was constantly late to everything because of this driving back and forth, to the point where her career was in a downspin and she thought she would have to quit and go on disability. She wasn’t able to go out with friends, she wasn’t even able to go to restaurants because she would keep fretting she left the hair dryer on at home and have to rush back. She’d seen countless psychiatrists, psychologists, and counselors, she’d done all sorts of therapy, she’d taken every medication in the book, and none of them had helped.

So she came to my hospital and was seen by a colleague of mine, who told her “Hey, have you thought about just bringing the hair dryer with you?”

And it worked.

She would be driving to work in the morning, and she’d start worrying she’d left the hair dryer on and it was going to burn down her house, and so she’d look at the seat next to her, and there would be the hair dryer, right there. And she only had the one hair dryer, which was now accounted for. So she would let out a sigh of relief and keep driving to work.

And approximately half the psychiatrists at my hospital thought this was absolutely scandalous, and This Is Not How One Treats Obsessive Compulsive Disorder, and what if it got out to the broader psychiatric community that instead of giving all of these high-tech medications and sophisticated therapies we were just telling people to put their hair dryers on the front seat of their car?

I, on the other hand, thought it was the best fricking story I had ever heard and the guy deserved a medal. Here’s someone who was totally untreatable by the normal methods, with a debilitating condition, and a drop-dead simple intervention that nobody else had thought of gave her her life back. If one day I open up my own psychiatric practice, I am half-seriously considering using a picture of a hair dryer as the logo, just to let everyone know where I stand on this issue.

Miyamoto Musashi is quoted as saying:

The primary thing when you take a sword in your hands is your intention to cut the enemy, whatever the means. Whenever you parry, hit, spring, strike or touch the enemy’s cutting sword, you must cut the enemy in the same movement. It is essential to attain this. If you think only of hitting, springing, striking or touching the enemy, you will not be able actually to cut him.

Likewise, the primary thing in psychiatry is to help the patient, whatever the means. Someone can concern-troll that the hair dryer technique leaves something to be desired in that it might have prevented the patient from seeking a more thorough cure that would prevent her from having to bring the hair dryer with her. But compared to the alternative of “nothing else works” it seems clearly superior.

Thanks to Richard Mlynarik for leading me to this, indirectly.  He actually pointed me to an interesting article about psychology and network theory:

The idea is that some mental disorders, instead of having a single "root cause", are a network of symptoms that reinforce each.  Some, not all!

Post has shared content
Great article...long but worth reading. One of the things they learned is the about the importance of "psychological safety" in any group

Post has shared content
> Welfare has three purposes that everybody seems to forget.

> 1. Welfare is about crime suppression. When people are starving, they will do anything and everything to feed their families. Theft, arson, and murder become easy choices when starvation is the other option. When the welfare system started, it is estimated that in some areas, the crime rate dropped by eighty percent.

> 2. Welfare is about stability. When a person becomes uninvested in the society he or she lives in, they begin to see the government as an enemy and society as a threat. People like Timothy McVeigh, Ted Kazinski, and Ammon Bundy are all persons uninvested in society. When significant numbers of people become convinced that the government is not interested in their survival or well-being, then revolution becomes possible. The American Revolution, the French Revolution, the Russian Revolution, the Chinese Revolution, and the Cuban Revolution all started when the underprivileged felt used and abused by the society they lived in. The U.S. keeps close track of the unemployment figures because if it reaches 10%, the crime rate rises. If it reaches 20%, the nation has become dangerously unstable. A nation rests on its lower classes. If they stop believing in the nation, woe be that nation.

> 3. Welfare protects the economy. In a recession, the money is locked up. The banks stop lending. The rich stop investing. The middle class scrimp and save. But the poor cannot afford to scrimp and save. The poor must spend every cent they can get their hands on just to survive. If a millionaire earns 10 million dollars, he/she will spend maybe 20 percent. The rest just goes into the bank, added to the pile. It offers little benefit to the economy at large. But if 10 million dollars is dumped into the welfare system, 98% is back in circulation within a week. Recessions happen. They can’t be helped. But a responsible government takes precautions and puts safeguards in place to minimize the damage that a recession might do. Welfare is one of those safeguards. It doesn’t matter whether the recipients of welfare deserve the money given to them. They are inadvertently serving a purpose in a big economic engine. What matters is that they spend the money and quickly. They are keeping the economy fluid and dynamic even in a recession.

-- Christopher Valdez,

Post has shared content
> The problem isn’t that [no psychiatric symptoms would ever cluster] together – depression, for example, is a very natural category. But so are various subtypes of depression. And so are various supertypes of depression, like depression + anxiety, or depression + psychosis, or depression + anxiety + psychosis. Choosing to draw the borders around depression and say “Yup, this is the Actual Disease” isn’t a bad choice, but it doesn’t jump out of the data either. When people try to use sophisticated clustering algorithms on psychiatric disorders, they usually come up with something like this, where there are only three supercategories instead of the 297 different diagnoses in the DSM. And even three supercategories are pushing it – people with psychosis are far more likely to have depression too! Having any number of categories starts seeming arbitrary and fuzzy.

> So Nuijten, Deserno, Cramer, and Borsboom (from here on: NDCB) ask: what if that’s wrong? What if there isn’t a latent variable like “influenza”? What if it’s symptoms all the way down?

> Consider a network in which each symptom is a node, connected to all the others by pathways with certain weights on each direction. So for example, “sleep disturbance” might be connected to “fatigue” by a strong path – people with disturbed sleep are much more likely to be tired. These might both be connected to “low mood” – people who don’t sleep well, or who are tired all the time, start feeling down about themselves. And this path might go the other way too: people who feel down about themselves might have more trouble getting to sleep on time. And maybe all of these are connected to suicidality, because if you feel bad about yourself you’re more likely to commit suicide, and if you’re suicidal you might feel bad about it, and if you’re tired all the time then maybe you can’t accomplish anything useful with your life and so death might seem like a good way out, and so on.

> Each node might affect the others with a certain delay. Being suicidal might make you feel guilty, but even if your last suicidal thought was fifteen minutes ago, you might still feel guilty now. Maybe it would take months or even years before you no longer felt guilty about your suicidal thoughts. So there could be loops: in a simple model, your low mood makes you feel suicidal, your suicidality makes you feel guilty, and your guilt makes you have low mood. This type of loopy network might be stable and self-reinforcing. Maybe your boss yells at you at work, which makes you have a bad mood. Then even if the direct effect of your boss would go away quickly, if it causes suicidal thoughts which cause guilt which cause more low mood, then the cycle can stick around forever.

> In NDCB’s model, all possible psychiatric symptoms are connected like this in a loose network. Particularly tight-knit symptom clusters that often active together and reinforce each other correspond to the well-known and well-delineated psychiatric diseases, like depression and schizophrenia. But there are no natural boundaries in the network; low mood and poor sleep may be closely connected to each other, but they’ll also be more distantly connected to anxiety, and even more distantly connected to psychosis. This corresponds to the fact that some depressed people will develop psychotic symptoms, even though psychosis isn’t usually associated with depression. The paths aren’t usually as strong as those between low mood and poor sleep, but they’re there, and in some people with a predisposition to psychosis or some idiosyncratic factor strengthening those paths beyond their usual level in the population, that will be enough.

Post has shared content
The news about Nazis in DC may have hidden some of the most important stories brewing about the new administration. This one may prove to have some of the most serious consequences. Trump wants to eliminate NASA's Earth Science division, one of the foremost institutions in the world studying the state of our planet. This team has key responsibilities in lofting the satellites which give us a view of what's happening around us – as well as being one of the world's best groups of climate modeling.

Eliminating the division would both disband one of the best such teams in the world, and eliminate nearly $2B of funding from the subject, costing hundreds (or more) of jobs across the field, and likely most affecting the careers of young researchers – with huge consequences for the field's future, just as it is becoming ever-more critical. (Remember that the Arab Spring was triggered by droughts in Asia and the Middle East, and the current massive surge in temperatures in the Arctic – it's currently 36°F above normal! – are going to have tremendous consequences. There's a reason the DoD considers this a top strategic priority.)

More chillingly still, the justification for this is that he calls their work "politicized science" – which is to say, "science which is politically inconvenient for him." We've seen a similar game in which Congress has banned any medical or epidemiological research on guns, because of the NRA's (probably justified) fear that the results of even the most trivial research would harm their political goals.* (Pro tip: if you know that anyone seriously looking at a question will come up with answers that hurt your goals, this may mean your goals are shady.)

There's actually a name for this sort of thing: Lysenkoism, named after Soviet agriculture director Trofim Lysenko. Lysenko considered genetics to be politically unfavorable to Communism, because if traits are inherent, then they can't be improved by the government, and that would make all sorts of parts of the Five-Year Plan obviously infeasible. With Stalin's enthusiastic support, all funding for research which disagreed with this was cut. (And this being Stalin, researchers who disagreed were shipped to Siberia) It was replaced by a rather bizarre official theory in which, for example, rye could be turned into wheat, and exposing wheat seeds to high humidity and low temperature would "teach" them and their descendants to grow in the winter.

The thing about science is that it's about asking questions and noting what's happening in the world around you. You don't get to make theories up and just say that the world is so; all you can do is describe what's actually observed, and try to figure out if you can predict what will happen next. That is, science is descriptive, it's not normative.

And that means that science is about things that keep happening, whether you believe in them or not. Unlike saying "I don't believe in fairies!," you can say "this wheat will grow in Siberia!" as often as you like, and the wheat still isn't going to grow there. That's the problem with Lysenkoism: it's based on pretending that nature works some way, and threatening anyone who dares to disagree with you, but nature doesn't really care. It will keep doing what it was doing before, and all that happens is that you've decided to be officially blind to it.

You do this with how crops work, and you end up with unexpected famines. You do this with how climate works, and you end up with unexpected droughts, floods, spreads of new diseases, and all sorts of fun and exciting things, because it turns out that the weather is still pretty important in our lives and you do not fuck with the laws of physics.

I say this with confidence: I was a physicist myself, am currently an engineer, and so if anyone is qualified to make a snappy answer to "Ye cannae the laws of physics, Jim!" it's probably me. But sorry: if your politics would be harmed by people being aware of reality, then all that means is that (a) your politics are apparently based on lying to people, and (b) at some point or another you are going to get a rude introduction to reality, which will not be good for either your politics or your constituents-slash-victims.

* Before anyone uses this as an excuse to go on a rant: I'm not anti-gun at all, and rather enjoy shooting. But there's a huge space between the Second Amendment and the sort of lunacy that the NRA has gotten infatuated with, where any restriction on a person's right to own a GBU-31 JDAM is tantamount to treason. And things like legally barring doctors from asking people if they have a gun in the house – even though, for example, that's a serious risk factor of death if anyone in the house is suffering from serious depression or similar illnesses – is just sacrificing human lives on the altar of their own political expediency. Seriously, fuck those guys.

Post has shared content
Today I learned that researchers at Jiao Tong University in Shanghai are applying machine learning to phrenology. I don't quite know how to say this with a straight face, but if your first response is to say that this sounds like applying ML to astrology but with more potentially sinister political motives, then you're probably thinking much the same thing as me.

Technical response follows, but the short version is that Chinese universities have become particularly infamous for encouraging faculty to publish absolutely anything, even pure nonsense, and this paper is entirely in keeping with that. If you reward quantity independent of quality, you get artificial intelligence crossed with phrenology for your pains.

The research paper, which is linked from the article below, is clean, elegantly written and illustrated, and a real showcase of how to combine p-fishing and inappropriate statistical tests to make utter nonsense look at first believable. What they have actually done is take a sample of 1,856 facial photos of Chinese men, about half of random people and half of criminals of various types (as specified by the local police, details not really clear), and built a pile of ML models until they found one which could reliably separate the two sets. Of course, each model has rather a lot of degrees of freedom (and the paper is mysteriously silent on just how many), which means they would have needed orders of magnitude more photos - at least hundreds of thousands, possibly millions depending on those numbers - to actually measure anything. It turns out that if you keep looking for "a combination of any number of facial features which describes all the people in set A but not set B," and you have more combinations of facial features than people, you can always find it.

They then "check" for overfitting by doing things like randomly relabeling the pictures and seeing that their model can no longer predict anything (of course it can't; their model was trained to identify the original sets it was told about), or feeding it pictures of Chinese women or Caucasian men and seeing that it can't predict anything. (No kidding; the sets of distinguishing facial features in those groups are different)

Contrary to the Intercept article, this paper isn't a failure of ML ethics; it's a failure of ordinary academic ethics, dressed up as machine learning. The paper is bullshit, plain and simple.

Post has shared content
It's time to get hands-on with A.I. Explore #aiexperiments and play with pictures, drawings, music, code, and more →
Animated Photo

Post has shared content

Post has shared content

Post has attachment
Wait while more posts are being loaded