Aretae

Feedback Systems Watch -- Education

noreply@blogger.com (Aretae) — Sat, 28 Jan 2012 01:27:00 +0000

Given that it's "School Choice Week", Education is a good stop for our next feedback systems analysis.

What systems of feedback do we need to watch in the education space?

Learner-Solution: The killer feedback loop is the feedback system between the student and the problem. All other solutions are subsidiary to this...but getting this going is tough.
Teacher-Topic: The 2nd feedback loop in the standard education model is between the teacher, the topic, and the learning outcomes. Do we change what or how we're teaching, based on how well the students learn?
Administrator-Outcomes: The 3rd feedback loop in the standard education model is between the administrator's choices, and the learning outcomes. The most recognized bit here is administrator's capability to fire teachers. But it's bigger than that. How much learning does $1000 of computer investment buy, as compared to an extra $1000 investing in teachers.
$-School: The 4th feedback loop in the standard model is when parents discover that the administrator is failing to educate their particular student, and move the student to a different school. Alternatively, the superintendent/state discovers that a school isn't working, and replaces the administrator.
Learning-Goal: A sometimes underappreciated feature of feedback loops is that they are primarily valuable when you have a goal. No goal...it's awful hard to tell if you're getting there. In the 1800s, the Germans developed the modern school system in order to teach obedience to soldiers and factory workers because the under-schooled did not have the requisite obedience to nominal authority figures...and their soldiers had lost badly to Napoleon. So...does the method of schooling that does not accomplish the goal of schooling get replaced?

How do they stack up?

Montessori is still the only big player in the learner-solution feedback loop. No one else seems to get it. Marginally, the DI folks...but not as clearly as Maria.
Teacher-topic? Most teachers do, very slowly. However, teaching experience doesn't get shared as much as it can. I think Kahn Academy works in this space.
Administrator-Outcome? I think it's illegal. The teacher's union is the big-evil here.
$-School. Fully broken. The purpose of school choice is to fix this feedback loop. It's a really good idea to have at least 1 working feedback system.
Learning-Goal: There is much hand-wringing, and NO action on this point.

Unsurprisingly, I support school choice. I even more strongly support bailling on the whole mid-1800s German obedience training for peasants model. But school choice (with the important point allowing schools and their associated teachers to fail/go out of business) would at least resurrect 1 of the feedback systems. Of course it's not THAT simple. So long as the district owns all the schools...and teachers impact learning quality (We estimate that a 2 sigma teacher is worth ~1 sigma of IQ)....and union rules require the district to keep long-tenured teachers on regardless of quality....you won't get massive improvement in-district. However, private schools will work to solve.

Another monster issue in education is that undert he current system, the feedback system to the parent is massively masked. IQ + Teacher Quality + School Quality + Peer effects + etc....it becomes very hard to tease real causal relationships out...and the individual parent is almost at the point of pure guesswork. Proxying student-happiness and Learning Outcomes are about as good as one can do.

And so I exhort: Support school choice. Private schools or homeschooling. Not only are you saving your own children, you are marginally fixing the feedback-free system, which helps everyone.

The Grand other-minds error

noreply@blogger.com (Aretae) — Fri, 27 Jan 2012 09:36:00 +0000

There is a way of thinking that is roughly universal among human beings...that also happens to be painfully wrong. I'm sure there are actually lots of these horrid things...but I want to focus on a relatively-smart-people problem today. The problem:

I am certain of this

Does not imply either of the following

It is true
It is clear to someone else with an even slightly different start point.

A 4th perspective on ethics

noreply@blogger.com (Aretae) — Thu, 26 Jan 2012 23:16:00 +0000

I've come out in the past as a relatively strong fan of (a) Jon Haidt's moral analysis (b) evo-psych analyses of ethics, and (c) David Schmidtz brand of back-referential meta-ethics. Yesterday, driving home from San Jose, I listened to a Russ-Roberts Econtalk podcast that gave me a 4th model of ethics: What ethics are necessary to have a society that improves the lives of its citizens?

In a very short analysis...here's how it goes:

1. In order for wealth, you need division of labor
2. If you have division of labor, you encounter (infinitely?) many situations where CEO has knowledge of option A, but Manager has knowledge of ascending options ABC, which the CEO cannot know about.
3. For division of labor to work, the CEO needs to expect that the Manager will use his knowledge of option C to better the company.
4. This doesn't work without real, justified trust in the other members of society.
5. Trust is complicated, and there are prerequisites to trust.
6. One prerequisite to trust is: Rule-based behavior, NOT Act-based behavior. (Rule utilitarianism, virtue ethics, and deontological ethics all pass this test. Act-utilitarianism notably doesn't)
7. Another prerequisite to trust is: negative rules trump positive rules.

Quick analysis...very interesting. I would think that my reactionary friends online would get a lot of mileage out of the position.

The Epistemology Problem

noreply@blogger.com (Aretae) — Thu, 26 Jan 2012 21:48:00 +0000

Most folks want to use epistemology as a way of justifying what they already "know" to be true. The challenge instead is to find an epistemology that lets us discover what is true, and allows us to update our beliefs when we find out that we were wrong. I argue that the best of these is an anti-certaintist inductive bayesian empiricism. Indeed, I think we are specifically scientifically justified in accepting empiricism, even though I am generally more than a little suspicious of using science to ground philosophy.

Observations:

Certainty is a feeling, which is not run by the brain subsystems responsible for logic. Rather, it's part of the same subsystem that handles love and anger.
Deductive conclusions outside of math have a hideous track record. No sane person could inductively conclude that deducing about the world (absent experimental evidence) is substantially more justifiable than astrology.
Induction, with Bayes, gives us NO way to reach logical certainty. Solution: bail on certainty.
People make LOTS of mistakes. Any epistemology with no error-correction mechanism is broken.
We know scientifically that the human logical mind develops as a goal-directed induction-engine.

Apostate Illka

noreply@blogger.com (Aretae) — Thu, 26 Jan 2012 17:35:00 +0000

Says Douglas Adams isn't funny. If I didn't agree, I'd have to post a screed about Illka's complete lack of taste.

Feedback systems watch -- Health Care

noreply@blogger.com (Aretae) — Thu, 26 Jan 2012 07:21:00 +0000

Aretae's 1st law is: The feedback system dominates.

Being as I'm among the loudest folks on the internet pushing this line...and because I've gotten some feedback (!!!) recently that suggests that this is one of the more valuable services that this blog provides...I'll be attempting to spend more energy looking at data from a feedback systems perspective.

Where are the feedback loops in Health Care, and which ones are working right?
The available feedback systems that I can see from here:

Individual choices, outcomes: Eat bread --> get fat. Exercise --> get healthy, but not thin. Take Vit.D in the morning and Sleep better that night.
Patient : Doctor system. Doctor says: Do X, Patient tries it, Doctor receives info about results
Doctor: Prescription system. Doctor suggests X...cumulative results cause success 30% of the time, therefore doctor changes pattern.
Patient: $ system. Patient spends money...solution doesn't work...patient doesn't spend money same way next time...and neither does his brother.
Doctor: $ system. Doctor who cures illness frequently gets paid better than doctor who doesn't...so the doctor is incented to do things right.
New treatments feedback system.

How many of these feedback loops are operational, powerful, effective.

Choice/Outcome: The problems with choice/outcome is twofold.

Choices only lead statistically to outcomes...many results are very hard to figure out...and the pattern-seeking brain generalizes to patterns even when there are none.
It takes work. Seth Roberts argues that the core feedback loop in Health Care is the individual and what he/she is doing. Seth-Roberts style 10-minutes of self-testing and record-keeping a day would do more to cure health issues in the world than any other set of health practices imagineable. 10 minutes a day is a lot.

Patient/Doctor: No feedback at all. Doctors don't make money from phone consults (because of medicare rules)...therefore there are none. Someone on my blogroll wrote about this in the last week. Because visiting the doctor is an ordeal...you don't have a normal-cases feedback system. Just an extraordinary cases system.
Doctor: Prescription -- Very low feedback. Doctor prescribes...no results unless it doesn't work badly.
Patient: $ -- Almost no feedback. Patient pays insurance company...insurance company pays doctor. Insurance company increases premiums next year. Interrupted feedback loops are atrocious.
Doctor: $ --Almost no feedback. Long term success and payment are unrelated.
New treatments -- Actively inhibited. Doctor Licensing and FDA approval form an almost inpenetrable barrier of innovation-resistance.

I'm counting 0/6.

Currently, there is some motion on 1B...which is great.

6 is the super-move...but no way no how is that going to get fixed. In the absence of 6, the best health-system known to man (Singapore) gets #4 correct, while our 2nd best option is to get #5 correct. A bad 3rd choice, still better than what we currently have is to replace #4 with state/$ single payer choice.

Any working feedback system, with a rapid rate of: bad results lead to abandoned practices, and new things get tried would be great. BUT...we have 6 available feedback systems and ALL of them are broken.

QoTD

noreply@blogger.com (Aretae) — Wed, 25 Jan 2012 17:10:00 +0000

Via insty: Mickey Kaus re: SOTU:

About a third of the way through I put it down and switched to reading the ingredients on the granola bar I was eating, just to perk myself up.

Funny oTD

noreply@blogger.com (Aretae) — Wed, 25 Jan 2012 05:01:00 +0000

Sonic Charmer:

Romney-Bain Matrix.
Eigenstate of the Union. (Math/Physics joke)

Endorsement

noreply@blogger.com (Aretae) — Tue, 24 Jan 2012 23:13:00 +0000

I hereby endorse Newt Gingrich to win the Republican nomination. (not predict...haven't figured that out yet...mostly expecting Romney still)

Almost certainly, if Newt wins the nomination, I will remain in my default anarchist state, and actively abstain from the complicity that is voting. However, the election will be SO much more entertaining with Newt as the candidate that IF you like spectacle, it would be insane not to endorse him.

If I had to vote...I'd probably vote Obama over Newt...for the best of the reasons that Borepatch gives for not supporting Romney...I don't want the Rethuglicans to pass their agenda either, and with Newt, they might.

OTOH...the positive case for Newt is that

FOUR of the supreme court justices are over 70...Ginsburg, Breyer, Kennedy, Scalia...something like a 4% chance of death on average per year per justice, over the next 8 years, given their ages.
the party that gets us out of this recession will have their economic ideas gain traction.
Newt is the only (non-zero probability) candidate standing that might do something actually crazy. Repeal the executive order allowing unionized federal employees. Unilaterally leave Iraq. Change the drug laws, or the FDA...push for an HSA-based Medical plan.

Interesting, related: Mead points out that the early-Abortion and gay-marriage debates are all but settled, in favor of the leftists...with (roughly) ALL young people in favor of early abortions and gay marriage. This leaves the leftists almost completely free of bogeymen with which to scare libertarian-leaners...like most of the youth in the country.

Orson Scott Card self-identifies as a conservative Mormon democrat, and endorses Newt (in the second half of the long article).

Jehu's article on Newt is also good.

More on Induction

noreply@blogger.com (Aretae) — Tue, 24 Jan 2012 22:36:00 +0000

How much of foundational epistemology does simply starting with induction solve? All of it?

If you start with induction...and conclude (very quickly) that deduction predicts well inside math, and rarely predicts outside of math...don't you just win out real fast?

You give up entirely on the notion of certainty...and you adhere pretty fast to the "map, not territory" model.

The scientific method falls out real fast as a how to purchase information from reality using induction .

Lean/iterative processes fall out almost as fast as you look for improvement methods.

I'm thinking that this is, while not necessarily shocking or new, the BIG win for epistemology.

Evolution cycles back around and supports the notion of the inductive mind...

Useful Thinks

noreply@blogger.com (Aretae) — Tue, 24 Jan 2012 20:16:00 +0000

Drezner against the doom-ists.
Caplan pdf pro-immigration. Key thought process: "To justifiably restrict migration, you need to overcome the moral presumption in favor of open borders" And then he shows that it's pretty hard to do so.
Caplan brilliantly skewers anti-trust from a novel perspective.
Isegoria: lecture sucks. I love how Isegoria finds all the research that says I'm right about education.

The Problem of Deduction

noreply@blogger.com (Aretae) — Tue, 24 Jan 2012 08:10:00 +0000

(Western) Philosophy from Thales through Aquinas proceeded peaceful and undisturbed over 2000 years. Then Descartes upset the entire endeavor by asking how do we know about the world? By attempting, and failing to answer...he opened up the field of inquiry. Hume's enquiry finished the job, utterly demolishing the pretense of knowledge, and most famously presenting the problem of induction...namely that if you try to justify induction by means of deduction, you cannot. 200 years of failed answers later, Hume seems to have won. But Hume wasn't radical enough.

I assert that philosophy has it 100% backwards. They attempt to justify induction using deduction. In fact, induction is the core process, and decution is a derivation. Outside of mathematics, induction has better predictivity than deduction. Properly, we ought not worry at all about the problem of induction. Induction, instead, is the base.

We should worry about the problem of deduction, and how deduction gets (partially) justified from induction, but almost universally fails in practice.

otDs

noreply@blogger.com (Aretae) — Tue, 24 Jan 2012 07:58:00 +0000

Caplan -- Schooling & marriage. Why do we treat them differently?
CDC attempts to transcend sanity.
McArdle on the future of retail.
Oz conservative gets QoTD:
I don't know, I went to a school that prides itself on free thinking and allowing students to make up their own mind, and it's general consensus that the white male has not only dominated the world the last 3 centuries, but that it has caused by far the most suffering against other people, both for women and other cultures and races
Group selection still non-predictive (pdf) HT: Evolutionary Psychology
Isegoria cites the "tutoring beats every other option" evidence. Feedback systems. Full stop.

Pic

noreply@blogger.com (Aretae) — Tue, 24 Jan 2012 06:55:00 +0000

Artsy wife, 2 very cute kids. Great picture.

Point about Chaos and Epistemology

noreply@blogger.com (Aretae) — Mon, 23 Jan 2012 07:56:00 +0000

Systems modeled mathematically frequently end up in non-linear dynamic systems...frequently referred to as Chaos or Chaotic systems. For those who are less than perfectly math-geeky...the super-short version of chaos is: EVEN given starting conditions (to almost any level of precision), and exact rules...there are some cases where you cannot predict the future, even a little ways out.

One standard example is the butterfly and the hurricane, where a butterfly flapping it's wing in Egypt results in a Hurricane that hits North Carolina 8 weeks later. Basically...the math works out so that incredibly small changes in initial conditions result in massively different final outcomes.

Another standard example is the Mandlebrot set ... a (very pretty) fractal measuring how many iterations a function can be applied to itself, starting with an initial value in the complex plane before its absolute value exceeds some value.

An awful lot of the math in nonlinear dynamics consists of finding those boundaries at which the system has large (infinite?) sensitivity to initial conditions.

I fear that in my discussions of epistemology and statistical prediction, I've neglected to mention (a) my awareness of chaotic systems...and (b) it's obvious relevance to my radical uncertainty.

Consider it mentioned.

Missing Comments

noreply@blogger.com (Aretae) — Mon, 23 Jan 2012 07:36:00 +0000

I've heard now from two separate folks (Wobbly, and Sonic Charmer) that they've been unable to leave comments. At the same time, I've had a bunch of folks leave comments in the last few days. My email address is listed to the right under my picture. If you've tried to leave a comment recently, but been rejected...please let me know.

Short summary of Aretae

noreply@blogger.com (Aretae) — Sun, 22 Jan 2012 20:35:00 +0000

We're all wrong a lot.
Goals come before truths. The good > The true
Induction (and statistics) is the only path to truth (predictivity). Those truths are statistical, and the variance is important

Good news: If induction teaches you that other choices work as well...that's legit.
Bad news: the only other path that works as usable truth is math...and it's only inside math.
Good news: Science gives you a regularized method for quickly determining most falsity
Bad news: Complex deductive logic fails the induction test.
Good news: The average opinion, (bet-weighted?) is a really cheap, good choice.
Good news: Atomic analysis is MUCH stronger than holistic analysis.

Feedback systems FTW.

Measure ONLY the speed and fidelity of the feedback loop...you know the future of the system

...and probably the present.
...for ALL systems
See: evolution, game theory, (micro-) economics, lean, agile, learning, ...

Evolution is the best scientific theory ever proposed
Lean is the best business mechanism ever proposed

Economic growth rate (per capita) is the only social metric that matters.

The first-glance atomic social metric is positive freedom.
Positive freedom REQUIRES wealth and negative freedom
People are willing to pay (a lot) for negative freedom
Wealth, therefore is the meta-1 metric.
Therefore the meta-2 uber-metric is economic growth rate.

Monkeybrains

Introspection doesn't tell us the same thing as watching other people
Each of us is basically like other people.
Simplest answer: introspection is wrong.
Understand other people via observation...apply to self.
Best current model:

evolved modular mind
decisions are explained by consciousness, not made by consciousness
heavy envy + sex components
conscious component evolved actively to be unaware of underlying system.
Buddhists would say: "the self is an illusion"

[Update] Wobbly suggests that I need this number to indicate: Summarize a lot.

PoTWeek?

noreply@blogger.com (Aretae) — Sun, 22 Jan 2012 18:44:00 +0000

Blunt Object finds it. RTWT.

Linkfarm

noreply@blogger.com (Aretae) — Sun, 22 Jan 2012 18:15:00 +0000

Sonic Charmer kindly responds (twice) to the Romer paper: Takeaway: "Is this even a thing that needs to be regulated at all? What problem is it solving?". RTWT.
Great link (NYT), commentary around offshoring the iphone by Cowen. Cowen is wrong, though (I say that very rarely). The primary value is company wants worker sufficiently poor to boss around.
Chris Christie just vaulted to favorite Governor with Gary-Johnson-esque sanity. Mead. WSJ.
Cowen finds useful thinking on Japan.
Balko's top 3 links on Gingrich, who he doesn't like, are hilarious.
HBDChick finds the most interesting discussion around evolutionary IQ I've seen in months. We already know (from prior Aretae noodlings) that heat-dissipation is one of the BIG problems in animals. We already know that heat dissipation is currently THE big problem in computer chips. What happens when we think about heat dissipation in the context of brains (HBD warning)?

Against Feser

noreply@blogger.com (Aretae) — Sat, 21 Jan 2012 20:16:00 +0000

On ImNotHerzog's recommendation...I've been reading the Aristotelian/Thomist scholar Edward Feser for between 6 and 15 months (my memory fails). I've finally reached an understanding...though after my conclusion, I went out and read some others...some of whom see the problem similarly to me.

I find Feser's critiques of folks like Rosenburg to be relatively solid...you can't play in philosophy by saying: science is all we need to know.

The traditional, Aristotle-Thomas-Rand-Feser notion of philosophy is: Metaphysics first. Then Epistemology. Then Ethics. Etc.

This approach is wrong, and no longer tenable. Question number ONE is "how do I go about deciding what to believe". You CANNOT, after Descartes, start from the place the great thinkers of the past start from. Thales, Pythagoras, Heraclitus, Parmenides, Plato, Aristotle, and Thomas all (IIRC) start from metaphysics. The problem is...once you've heard the questions of Descartes and Hume you can't start (in philosophy) from metaphysics any more. Epistemology is the central question. Everything else comes after.

Feser seems to be the pre-eminent modern defender of the metaphysics-first approach that Aristotle takes. If he'd been writing 400 (even 362) years ago...this would have been a marvelous way of thinking...and he'd have been considered brilliant. Now...it's as if he's writing about health, but missing germ theory. Sure, you can write very intelligently about health missing germ theory...but not for very long.

I have not yet read Feser's book. But even some of the sympathetic reviews (Feser's response) I've read suggest that my analysis above is pretty solid, and I'm likely to encounter the same problem in his book.

Summary position: Feser starts wrong in philosophy. Rand, with her superb ability to identify problems with opposing positions (much less woth her own) would have correctly called his approach "intrisicist", as opposed to the also-wrong "subjectivist" or correct "objectivist". While I find the distinction terribly important, and correctly leveled...I think that the problem is earlier. Deductive logic is wrong SO often that one cannot rely upon it as a path to truth without feedback systems. Indeed...even if you don't take the hyper-radical Aretaevian skepticism route...you're still left with the questions of epistemology...and the inductive challenge to deductive logic: You're wrong a lot, even when you sound right. Why should I believe you this time?

Philosophers would love to say that the truth is evident...but my feedback system says no. The feedback system on deductive logic applied to ANYTHING outside of math/formal logic is atrocious, and it's results are at least as bad. [quibbles about simple physics like orbital calculations ignored].

Reading Differences

noreply@blogger.com (Aretae) — Sat, 21 Jan 2012 18:39:00 +0000

I have been reading other folks critiques of works I am very fond of, and I'm somewhat baffled.

Here, Edward Feser discusses Dennett in an aside, and suggests: "Daniel Dennett is one naturalist who does not see this, or at least who constantly helps himself to teleological concepts which he cannot successfully “cash out” in naturalistic or non-teleological term". I consider Dennett's book "Darwin's Dangerous Idea" among the most important books written...and to a significant extent, the book is about how natural english talks in teleological terms, but that's not what's going on. "Helping himself to teleological concepts" is very distinctly NOT something that I remember to have occurred in the book.
Here, Robin Hanson mildly disses my new favorite book. Also, slightly, here. My read of the book seems very different than that of Hanson. I read one of the major theses of the book to be that the mind works better if there is intentional non-coordination. Kurzan explicitly argues that a press secretary who actually knows what's really going on can't lie as effectively as one kept in the dark...and the press-secretary (consciouness) mind can't lie (be hypocritical) as well if was fully informed. Hence, hypocrisy is a substantial portion of the purpose of consciousness.

Why the heck are the readings so different?

Useful Links

noreply@blogger.com (Aretae) — Sat, 21 Jan 2012 18:00:00 +0000

Bryan Caplan checks the age-adjusted return on sheepskins. Fascinating. As someone who does a reasonable amount of interviewing/hiring in the IT industry...I can say that the amount of weight I put on a college degree is precisely zero. Usually, I don't read the resume that far down. Years of experience * conversation measured Tech-IQ. Further and further, the signaling (-only?) model of education seems overwhelming.
Envy uber alles: Henderson finds a paper that measures the laffer curve on capital gains taxes. Apparently, counting ONLY capital gains (not secondary effects), the curve Maximum is just under 10%. RTWT.
David Brin has a great BIG list of good sci-fi. his list is far better than the Reader's Digest one...what with being a sci-fi author and all.
Robin Hanson praising religion. As per my recent fights with ImNotHerzog...it's pretty clear that religious people are happier. I'm moderately convinced that it's MOSTLY community + gratitude effects. I'm also convinced that for maximum hedonic benefit from religion, one should be either Mormon or Buddhist...with Mormonism maximizing near-term happiness for relatively low life cost (silly beliefs about magic underwear and sky gods), and Buddhism's meditation practices maximizing long-term joy at relatively high life cost (compassion meditation practice for 20 years). A goals-first kinda guy like me doesn't dispute the utility of religion...just the truthiness.
Casanocha asks the question -- How do you deal with high talent folks who very simply will have other, better opportunities in the future. Great think.
Isegoria finds important numbers: The real job market is informal, with no resumes, and only personal recommendations. At 80% of the market.

The modern problem

noreply@blogger.com (Aretae) — Fri, 20 Jan 2012 21:37:00 +0000

All socioeconomic problems (+/- 3 %) are caused by too little patience....too much impulsivity.
All reproductive success problems among the upper classes are caused by too much patience...too little impulsivity.

Is it that simple?

Mockery is the best Medicine

noreply@blogger.com (Aretae) — Fri, 20 Jan 2012 18:16:00 +0000

"Cougars" is an insensitive mascot name ... because of the obvious reason? I think they should rename themselves the gold-diggers.

PoTD

noreply@blogger.com (Aretae) — Fri, 20 Jan 2012 18:11:00 +0000

Bloody Shovel is dangerously close to joining Robin Hanson on my list of -- don't link to, because everyone with an IQ over room temperature (celcius) -- should be reading him.

Again, with the fiction of the nation.