News for hackers. Stuff that matters.
Re:A ridiculous Common Core test for first graders by ghshephard on Friday November 01, 2013 @11:40PM (#6658054)
Question 1 was bizarre. I'm not even sure what they were trying to ask.

Question 2 "There are 8 Jars. 6 Jars have Jelly. The rest have peanut butter. How many Jars have Peanut Butter."

The first thing that occurred to me is that we didn't specify the Jars with Jelly don't have peanut butter. Horrible question.

Question 3,4 I can live with.

Question 10 was greek.

Question 11 - Once again, nothing states that the others aren't red as well. Reinforcing poor logical thinking.

Question 12 "Related Subtraction Sentence?" No wonder these kids are crying.

I couldn't read any further. This is how they are trying to teach kids math?

Re:A ridiculous Common Core test for first graders by harshreality on Saturday November 02, 2013 @01:05AM (#6658233)
The test becomes much simpler after realizing that it covers simple addition and subtraction and that it was written by someone who failed logic 101 and doesn't know who Edward Tufte is.

When a question doesn't make sense, visual positioning of the elements offers clues, and in the logically incorrect questions, there are simple assumptions (only two red in question 11, only pb or only jelly in question 2) that make the question a simple addition or subtraction problem.

I'm not saying the test is good for first graders, but I think complaints that problems are wrong or confusing miss the point of the test. It is not a logic test.

Re:A ridiculous Common Core test for first graders by Jach on Saturday November 02, 2013 @01:06AM (#6658237)
I'm okay with questions 2 and 11. Admittedly #2 could be made more explicit by adding the word "only" after "have", but I think in the absence of the word "some" then it's a valid inference to conclude the writer meant "only". Same thing with #11, it could have been made more explicit (in two ways!) by saying "only 2 are entirely red" and asking how many are not entirely red. But again, I think in the absence of a phrase like "at least two of them are partially red" then inferring "only two are entirely red" is valid. I also think the majority would have no problem making those inferences -- the person whose test was uploaded didn't have a problem (they just failed with subtraction on #11).

I think I get #5. The person wrote a 9 when they should have written a 4.

#6 is fine, but then it makes #12 not make sense. Shouldn't a subtraction sentence have a subtraction symbol in it? Or is a "related subtraction sentence" an actual thing, not just asking what the subtraction sentence is in relation to the picture?

#7 is fine.

#8 introduces the "number sentence". (What's wrong with the word "equation"?) But aren't all of the examples "subtraction sentences"? Bah! Apart from that it's fine...

#9 introduces the "subtraction story". Guh?

Re:A ridiculous Common Core test for first graders by ghshephard on Saturday November 02, 2013 @06:27AM (#6658762)
"but I think in the absence of the word "some" then it's a valid inference to conclude the writer meant "only""

The first two years of critical thinking were dedicated to trying to drive that type of logic out of our head. "Six Jars Have Jelly" in no way implies they don't also have peanut butter, it simply makes explicit that six jars have Jelly. "The rest have peanut butter" doesn't imply the rest don't have jelly, it simply makes explicit there are at least two jars with Peanut Butter. That entire sentence would have been perfectly applied to 8 jars of peanut butter/jelly. By suggesting it does not, you are setting in place logical frameworks in a child's head that become very difficult to remove - even if it's clear they are invalid frameworks.

When you say, "I also think the majority would have no problem making those inferences" - you are just making my point - the Majority of people come to conclusions that are in no way supported by the text, and, even worse, when you point out their error, they instead suggest an inference has been made in the text, when it very clearly has not.

"Some of the Apples in the barrel" in no way suggests not all of the Apples are rotten. "A few people in the crowd acted out against the police" doesn't imply they didn't all act against the police, etc...

The person whose test was uploaded is a six year old who is doing their level best to come up with the "right" answer. In ten years, they are they same type of student, who, in a chemistry exam, when provided with the question, "What particle in an atom has the least negative charge" will answer, "Electron".

Re:A ridiculous Common Core test for first graders by Jach on Saturday November 02, 2013 @10:08AM (#6659223)
I wonder how many years of probabilistic thinking it would take to drive it back into your head?

What's your prior for p1 = P(jars contain jelly and something else | you're told "six jars have jelly"), and your prior for p2 = P(jars contain only jelly | you're told "six jars have jelly")? Are they the same for you? For me, they're quite different: p1 << p2. Why is that? As a first guess, I'm very unused to seeing jars containing jelly plus other crap, even though I have seen those strange peanut butter + jelly combo jars. (http://www.americansweets.co.uk/ekmps/shops/statesidecandy/i...) Furthermore I'm unused to being told "a jar contains X" when a jar in fact contains "X and Y (and Z and ...)".

The Majority use probabilistic reasoning (even if they often suck at it) in every day life. Rational agents (which humans approximate) use probabilistic reasoning, see Cox's Theorem and the work of others. This is (part of) why calling out a logical fallacy in an argument is often a mistake -- a lot of arguments aren't small (or large) trades of deductive proofs, they're probabilistic arguments where each side is trying to update the other side with supposedly new information. Many logical fallacies can turn into probabilistic theorems. For example, "arguments from authority" can become valid in a probabilistic framework. How? While [expert said X is true] --> [X is true] is a fallacious deduction (unless you assume as a premise that whatever expert says is true, then it's just a tautology...), probabilistically our everyday experience suggests P(x is true | expert said x is true) > P(x is true | random nobody said x is true). We can find plenty of cases where experts are wrong, but in general that pattern seems to hold, and we're more willing to accept counter-intuitive truths from experts than from people on the street.

To address one of your examples: [We're informed 'a few people in the crowd acted out against the police'] --> [Not all of them acted out] is indeed an invalid logical deduction without more premises, but let's treat this probabilistically. (Fair warning: I tried my best to make sure there are no errors in my math but I've been up all night writing signal processing code...)

Let x="Not all of the crowd acted out",

let y="We're informed 'a few people in the crowd acted out against the police'",

and let c=our general background context="We know an incident occurred with the crowd and the police".

We have a prior belief about x: P(x | c) = P("Not all of the crowd acted out" | "we know an incident occurred").

We receive a new piece of information, y: "We're informed 'a few people in the crowd acted out against the police'". Now let's perform a Bayesian update:

P(x | y and c) = P(x | c) * P(y | x and c) / P(y | c)

Clearly if P(y | x and c) > P(y | c), then our updated belief about "Not all of the crowd acted out" will go up, and the inference [y] --> [x] is valid. (You would interpret the --> not as "therefore" but "increases our confidence in". [Being informed the phrase 'a few people acted out'] increases our confidence in ['not all people acted out'].)

I think it's clear that P(y | x and c) is in fact greater than P(y | c). If you don't, then continue with this next section. Using marginalization we can break down P(y | c) independent of x, then use the product rule to break it down even further: P(y | c) = P(y and x | c) + P(y and not(x) | c) = P(y | x and c) * P(x | c) + P(y | not(x) and c) * P(not(x) | c)

So, is P(y | x and c) >? P(y | x and c) * P(x | c) + P(y | not(x) and c) * P(not(x) | c), subtract and factor:

P(y | x and c) * (1 - P(x | c)) >? P(y | not(x) and c) * P(not(x) | c), by the sum rule:

P(y | x and c) * p(not(x) | c) >? P(y | not(x) and c) * P(not(x) | c)

P(y | x and c) >? P(y | not(x) and c)

Reminding what all the variables are again (temporarily dropping off the c for brevity):

P("told 'a few' acted out" | "not all acted out") >? P("told 'a few' acted out" | "all acted out!")

I don't know about you, but I think it's very unlikely to be told 'a few' acted out if in reality everyone acted out. QED.~

To put my whole comment another way: try writing an AI and sick it on this test. If it doesn't use probabilistic algorithms and instead insists when it comes to question #2 that "I cannot know the answer other than it's between 0 and 8, the question isn't explicit enough!", it's going to perform very poorly. Human language is not crisp, treating it as if it were is a bad idea, that's why we use crisp math when we want to be crisp and if we can do automated proofs so much the better.

On your last remark, I find it amusing that if I take your "it's not explicit enough!" complaint seriously, I'm not sure whether you think the correct answer is supposed to be "proton" or "neutron" or "electron". That's because I'm not sure whether you mean "least negative" to mean to "furthest to the right from negative epsilon (i.e. 'most positive')", implying the proton, or "smallest member of the negative reals excluding 0", implying the electron, or "smallest member of the negative reals and 0", implying the neutron. If I wrote the question I think I would have meant "proton" as the answer. So given that the answer is proton, however, the error the student most likely made is just a parse error, failing to properly change "least negative" into "most positive". Similar to answering "yes" to the question "the sky is not not green?", or asserting -3 * -3=-9. It's a different sort of trick on the student than what you insist question #2 could be tricking the student. In #2, the trick would be "Aha, but we never said the jars were entirely jelly! Don't insert words that aren't there!" In this charge question, the trick is "Aha, you didn't parse 'least negative' into 'most positive'! Read carefully!"

Re:A ridiculous Common Core test for first graders by ghshephard on Sunday November 03, 2013 @01:29AM (#6662430)
I don't disagree with you, that, if we have to make a "Best Guess", we need to bring in our day to day context - but using the the one word, "only" - not only eliminates the guessing, it also helps to reinforce logical thinking.

Re: Electron - I was trying to use it as an example of a poorly written exam. This particular question was on a Grade 10 final exam, and I thought it was obviously, "proton." The "correct" (for some definition of correct?) was "electron." I was using it as an example to demonstrated that "test creators" - are people, just like you and me, who make mistakes. Also using it to show how the student who is always looking for the "correct" answer (as compared to the correct answer) starts to lose their ability to parse and logically communicate.

Re:A ridiculous Common Core test for first graders by mturmon on Saturday November 02, 2013 @01:37AM (#6658292)
"... #12 [does] not make sense. Shouldn't a subtraction sentence have a subtraction symbol in it?"

I have a kid who went through this curriculum, and yes, a subtraction sentence needs a "-" sign in it. The idea of a "related subtraction sentence" is a standard thing in this curriculum. I think #12 is the worst question on the test, because it is flat-out wrong.

Like I wrote elsewhere, I'm ok with "number sentence" because of the parallel with regular sentences. It gets the kids familiar with the idea that the math is making assertions.

I TA'd several undergrad engineering classes (at a top-10 engineering school) and you'd be surprised how many students did not use the equal sign in their work. They would write one expression down, and then another under it, and another beside that one, etc., but without equal signs, it's just babbling.

Re:A ridiculous Common Core test for first graders by seniorsassycat on Saturday November 02, 2013 @12:54AM (#6658200)
I don't agree with your complaints about 2 or 11, assumptions are a part of everyday communication. Anyway, wouldn't the question would be unanswerable if some of the jelly jars could have peanut butter? We know the question is answerable, therefor no jelly jars can have peanut butter.
Re:A ridiculous Common Core test for first graders by ghshephard on Saturday November 02, 2013 @06:39AM (#6658790)
Could you accept that these assumption are frequently made incorrectly?

Let us say I went shopping, and I know that I purchased eight jars of sandwich spread. I know that six of the jars had a mixed peanut butter/jelly stuff that you can squeeze out, and I remember the other two had peanut butter in a glass jar, but I can't remember if they were mixed or not, and I say, of those eight jars, "Six of the jars had jelly, the rest had peanut butter" - because that's all I was certain of - that somebody then inferring that the six jars did NOT have peanut butter, and that the other two jars did NOT have jelly -would be wrong in doing so?

It's entirely possible that the question was poorly put together, and the answers provided were not sufficient to correctly answer the question.

Test makers aren't all perfect beings - as clearly demonstrated by this test.

Re:A ridiculous Common Core test for first graders by glenra on Saturday November 02, 2013 @03:40AM (#6658481)
> Anyway, wouldn't the question would be unanswerable if some of the jelly jars could have peanut butter?

No, the question is answerable but it's answerable with a range rather than a specific number.

> We know the question is answerable

How do we know that? Lots of real-world questions aren't answerable and even if the author meant this one to be, textbook authors are fallible. So I don't think that's a fair assumption.

And yeah, I nitpicked exactly like that back when I was in grade school too. :-)