The reality is that many of the schools, and many of the teachers, are just bad. This can't be acknowledged in the public debate, so it isn't addressed. The unions have established a structure dedicated to protecting bad and mediocre teachers. Which is why great teachers can't get paid appropriately -- the unions would never allow it. It's also why teachers don't enjoy higher status: As a group, they really don't deserve it.

Yes, the unions are right that, without tenure and their workplace rules, teacher hiring and pay would be a political game. That's what political systems do, they respond to political incentives. That's the whole idea of charters: to move the accountability of the schools out of the political system and into the hands of a constituency invested in actual quality. It's probably true that _today_ parents as a group aren't completely equipped to handle that role, because they've been trained to just send their kids off to school. But as they get the opportunity to make more decisions and influence the schools' offerings, they'll come up to speed.

tldr; The testing apparatus is a symptom of a broken governance apparatus. The solution is to fix the governance and incentives. So long as the incentives are political, the results will be political.

As a population, on average, teachers simply aren't that impressive. Teaching is a noble vocation and ought to have great status. But when the average teacher simply isn't that good, people aren't going to extend much respect to the profession.

And the unions contribute to that condition by resisting efforts to improve the quality of the teacher population.

To extend that to the teaching population is madness. Where are you determing "the average teacher simply isn't that good"? How are you making that conclusion? Because it sounds like you just pulled it out of thin air, or it's just a personal view you have.

To say that teachers don't deserve a higher status because "some [imaginative] proportion" are "simply not good [according to you]" is laughable.

>Because a ton of scared people think the tests are the safety net of education. They are the last resort. If you got a bad education and can still past the test, it's a quantifiable bare minimum.

Why does test-driven development enable the efficient creation of software that works, but the same isn't true of education? For example, asking a student to read some text and answer questions about whether they've understood it is a simple way to see if they're actually literate. An assert() if you will.

Test-driven education has resulted in an American literacy rate of 99% (http://en.wikipedia.org/wiki/List_of_countries_by_literacy_r...

The SAT is a test that heavily influences University admissions in the United States. (Not all countries have college admissions exams.) The most selective universities in America are some of the ones who give it the greatest weight. It is hard to argue against a causal link given the level of scholarship and research that is produced in American higher education.

http://www.timeshighereducation.co.uk/world-university-ranki...

I am not going to weigh in on what kind of tests should exist. But the idea that testing is a last resort, and you can 'get a bad education' and still pass the test, is like the idea of a function that doesn't work, but passes all the tests. Not to put words in your mouth, but in essence then you are essentially arguing that it is impossible to test an education.

If so, what makes you think a bad education exists? Perhaps all educations are perfect?

Granted this last example is a bit glib of me, since perhaps there are things that are infeasible to test in a normal test setting. But overall, I would say the type of things that make you say someone got a poor education, are also the kinds of things that can be tested for and are, in fact, tested for. You don't mean they can't make an omelette without getting it all over the floor.

You can reduce the delta between passing tests by writing more, better tests, but again you quickly run into a problem - you spend so much time on testing that you start to lose the time to actually learn / write the actual project code.

So that's some ways in which the analogy is good. But there are some glaring errors in the analogy. In education, a student can have perfectly mastered the material, but is just really bad at passing tests, getting stressed out by it. In TDD, you can make your test environment match the target environment very closely. For most student exams this isn't true - most applications of higher learning are not done in a race against the clock, and you always have access to the Internet to help you out when you're stuck. By their nature, tests can't validate your capacity to work in a team with any fidelity. So, some of the very important skills that are needed in the application of knowledge are very difficult to test, making tests less valuable as a means of judging a student's acquisition of skills. There's a reason why PhDs require a thesis and a thesis defence, rather than sitting an exam before they are awarded.

However, because so much weight is put on testing, that is essentially what happens in school a lot of the time. For some reason, it seems more obviously absurd in the software context.

Do the job, make the thing, etc. If you have to observe the process.

On the other hand, TDD is formative testing: You iterate over the code until all the tests pass in a structured way. You don't fire programmers for writing code that doesn't pass all the tests on the first attempt, but deliver actionable feedback at every step. That kind of testing is massively useful in the classroom and teachers practice it with formal and informal assessments all the time.

I have a hard time reconciling this 99% figure with other statistics about illiteracy in the US.

http://www.statisticbrain.com/number-of-american-adults-who-...

14% of American adults can't read. 21% of American adults read below a 5th grade level.

The difference is partly the definition used. The CIA world factbook (the source of the WP information) uses "age 15 and over can read and write." The CIA World Factbook uses an estimate for the US 99% figure and does not give any explanation how they get that figure.

It's an estimate from census data. There are problems with that estimation.

https://en.wikipedia.org/wiki/Literacy_in_the_United_States

> The Census Bureau reported literacy rates of 86% based on personal interviews of a relatively small portion of the population and on written responses to Census Bureau mailings. They also considered individuals literate if they simply stated that they could read and write, and made the assumption that anyone with a fifth grade education had at least an 80% chance of being literate.

That feels like a pretty big flaw when we know that rates of literacy for 8th graders are below 50%. And the numbers are far worse when we look at math.

You are right: testing is important. The problems with testing are bad teachers can teach only to the test and lessons that could be about general problem solving approaches turn into "guess which answer is more likely from the multiple choice selection" - a less useful skill.

The other problem with tests is explaining the tests to adults. Two examples here. i) the extensive discussion about common core math. ii) England teaches "synthetic phonics". To make sure pupils are using real synthetic phonics to read and not just learning whole words the tests use nonsense words. This tests a child's ability to sound out the letters and then smoosh those sounds together. This caused some confusion in parents.

When it comes to testing teacher/school performance through its students, if you reconsider a test framework is given a body of objects:

* instantiated outside its control(students are brought, not created at school)

* instantiated with unique and unknown inputs(vs given known inputs)

* have hidden independent variables(personal traits)

* have hidden linked/shared variables(external traits, economic environment, local crime, family etc)

* have _few known inputs_ (teachers, school staff, grounds)

and then ONLY test pass/fail based on those _few known inputs_ and the results from each individual object, you will learn that the inputs pass/failed, but determining what the cause is will be far harder(strong linked variable or teacher? intelligence sapping virus?). unfortunately the frameworks executors believe that the tests clearly show that given the inputs they control, the results are repeatable.

I don't think testing a body of objects-- err --students is pointless, but directly linking their results to teachers and schools discounts so much more than this testing framework can account for. They should be a starting place for conversation, but not a litmus test.

> in essence then you are essentially arguing that it is impossible to test an education.

Education can teach techniques, such as multiplication table or spellings, and that part of education is easy to test. However, education also builds the foundation---of learning how to study, or of how to educate yourself, or of how to be intellectually responsible---and that part is hard to test. The effect might be only visible years after you leave the school, after you go through tough problems as a responsible adult.

I'd love the test-based quantitative approach, and I'm sure there are researches, but this required time span to run test-implement cycle seems to make it very difficult issue to tackle.

`assert(complexFunction() == 42)`

then the easiest way to make the test pass is just to return 42. The analogy in education is "teaching to the test". You can get people to recall small specific bits of information for the test without it actually being useful.

What do you test - memorizing or understanding?

Does it test the required minimum, then passing it corresponds to the lowest positive grade (whatever that is in your school system).

Or do you also test "nice-to-have" knowledge? Then why do you test it?

But that is true for all countries, while the pervasive test culture is a phenomenon mostly observable in the US. The quality of teachers (or their education) can't be the major reason for the test dependency of the educational system in the US.

A culture of forcing children to study for hours every night at home and go to school 6 days a week?

* Greatly increase teacher pay. Increase it to something like 1.5x or 2x the current pay scale.

* Give greater flexibility to the administration to fire underperforming teachers.

* Install much more robust (and flexible to change over time) measures of teacher success. The Bill and Melinda Gates Foundation has funded several studies in this area (https://docs.gatesfoundation.org/Documents/empowering-effect...).