Wednesday, February 17, 2010

Test if You Know What You Don't Know

One of my favorite quotes is due to the American humorist Josh Billings (Henry Wheeler Shaw):

"I honestly beleave it iz better tew know nothing than two know what ain't so."

Sometimes this is reported as:

"It's not what you don't know that scares me; it's what you know for sure that just ain't so."

Now here's a little test that claims to identify your "risk intelligence quotient": not what you know, but how accurate your estimates are of what you do know.

I'd love to see the blustering creationists Denyse O'Leary and Kirk Durston take this test. My guess is, they wouldn't score very high.

So take it, and leave your score in the comments.

46 comments:

Jimmy said...

I got a 72...embarrassed about a couple of my answers, but overall not bad.

Barry said...

65

You should tell us your score eventually, after the bulk of your readers' scores are in.

przemek said...

76; I'm rather disappointed. I guess I am overconfident in my ability to spot my own overconfidence.

Jeff said...

I came in with a 73. Is that good? My identity and calibration lines were both straight with similar slopes. The identity line started under the calibration and then went over the top of it after the mid-point. Not really sure what that means. I think it might mean that I have cancer, or that aliens are attempting to change the size of Africa...

One Brow said...

I had an 86.

Bayesian Bouffant, FCD said...

Mr. Google and I got an RQ score of 98.

The quiz has some ambiguous questions and questionable answers.

#9: "Cats are not mentioned in the Bible"

This is supposedly true, and yet look up EpJer.1:21
"Upon their bodies and heads sit bats, swallows, and birds, and the cats also."
(King James version)

Not to mention numerous mentions of lions. Are lions considered cats for the purpose of answering this question or not?

#15: "Humphrey Bogart had two wives before Lauren Bacall"

Bogart had three wives before Bacall. It is unclear as to whether the question is to be interpreted as asking whether Bogie had exactly two wives, or whether he had at least two wives"?

#28: "Natural gas has an odor"

I understand that methane does not naturally have an odor, but by the time it gets to my house it does. And yet the purported answer is "False."

If the quiz authors cannot properly vet whether their answers are correct or not, their results will be distorted.

D. Swart said...

88! Fun!

I think the height of the curve shows the fraction of correct answers that you marked with the confidence shown in the x-axis.

Of this I am 90% sure.

Anonymous said...

Bayesian Bouffant, FCD:

Is there no such thing as an odourless substance then? I can put salt in water, does that make water no longer tasteless?

Jeff: The identity line is everyones identity lind. Regarding the over under trend I read that to mean you're underselling yourself, (you're less confident than your accuracy shows you should be)

isohedral said...

66 -- average, I suppose.

Jimmy said...

Bayesian Bouffant, FCD:

I think the ambiguity is intentional. It definitely affected my confidence. i.e. If for example I knew that Bogart had 3 wives, but the question said 2. Then my confidence in a true answer would be less than 100% because I don't know if it is asking >= 2 or exactly 2.

Unknown said...

80. So-so, probably could've done better, had I taken more time, but I gather that the whole idea is to see if you know you know or know that you don't know.

Jonathan Lubin said...

Hmmm. I got 89.

Harriet said...

98 for me.

Bayesian Bouffant, FCD said...

Anonymous: Is there no such thing as an odourless substance then? I can put salt in water, does that make water no longer tasteless?

The question is, what do they mean by "natural gas"? If they had said methane, I would agree that it is odourless. But does "natural gas" mean what they pump out of the ground, or does it mean what they pipe to my house?

As for your example, WT*? Does water routinely come out of the tap, or get sold in bottles, with salt added? Do you think you have a point?

What would you say if I insisted that Hostess Twinkies do not contain preservatives, it's just that perservatives were added to them before they were put on sale?

---

Jimmy: I think the ambiguity is intentional...

Well then **** them. The purported purpose is to cross-correlate truth/falsity with confidence. You cannot do that properly if you are WRONG about assigning truth/falsity.

Imagine a roomful of Catholic Bible experts who were absolutely certain that the Bible mentioned cats, and were actually holding their Bibles open to EpJer as they took the quiz. They would get marked down as holding strong certainty for a wrong answer.

Do not defend the indefensible, it makes me think less of you.

cody said...

I took this test a little while back, I think PZ Myers posted it on Pharyngula, or at least I saw it somewhere. I got a 92, but in the scoring page it makes it sound like you should retake it, so the second time I got a 78, which was weird since I could remember a few I had gotten wrong, so it was easy to know the right answer.

For some time now I have been interested in creating a set of similar tests that could reveal how well (or not) two opposing groups understand one another's position. Specifically christians and atheists, or creationists/IDers and anyone with a brain. Or the mainstream republican party and an adult.

athel said...

82.

I agree that the ambiguity of some questions was deliberate. I didn't know (and don't care) whether cats are mentioned in the Bible or how many wives Humphrey Bogart had, and in such cases one is, I presume, supposed to choose 50%.

Filipe Calvario (from Brazil) said...

81. I'll wait for one in Portuguese, and see if I can score better.

Gary Will said...

Got a 95 taking it legitimately. Thought it might be more like a test of risk aversion so I tried it again answering 50% to everything I wasn't certain of and either 0 or 100 to everything else. That took it to 97.

Joel Noche said...

Ha! I'll bet no one can beat my score! I got a score of 1! (Yes, one.) Honest! I think I got around half of them correctly (I know this because they provide feedback after you answer each question), but there was a connection error when I clicked the "finished" button, so I had to refresh the page. The graph that came out had a slope of zero passing through the origin (practically, the positive x-axis).

I actually thought the website was a hoax (that is, you'll get a score of 1 regardless of your answers) until I read these comments.

Joshua said...

84.

And I strongly agree with the cats one. That was obnoxious. I don't know what they were thinking for that one.

Rachel said...

I got an 86.

Ipmilat said...

81. I know that I know very little about anything.

Anonymous said...

I got a 77. Not bad for a hard-headed blowhard that I am.

Miranda said...

"Now here's a little test that claims to identify your "risk intelligence quotient": not what you know, but how accurate your estimates are of what you do know."

It seems that part of the test is testing on what you know. Isn't it?

"I'd love to see the blustering creationists Denyse O'Leary and Kirk Durston take this test. My guess is, they wouldn't score very high."

Why not? Because they're not up do date on Humphrey Bogart's and L.L. Cool J's love lives? Slam these guys when it's appropriate, but don't make it seem gratuitous.

Bayesian Bouffant, FCD said...

I think the ambiguity is intentional. It definitely affected my confidence...

If so, then the confidence I have in my answers is not just in whether I have it, but in whether the quiz authors got it right. I.e. they are not measuring what they think they are measuring.

Jeffrey Shallit said...

It seems that part of the test is testing on what you know. Isn't it?

Not much at all - at least not if their description of the test is correct.

Because they're not up do date on Humphrey Bogart's and L.L. Cool J's love lives?

No. I know nothing about these things either, but I still scored high on the test. That's the point - the test is ostensibly not measuring what you know, but whether you know that you know.

Slam these guys when it's appropriate, but don't make it seem gratuitous.

No offense, but you've entirely missed the point of the test and my comment.

Barry said...

How about a test that measures how willing people are to change their wrong answer when confronted by real data that contradicts it?
After seeing a map that clearly shows the borders of Armenia and El Salvador, which people will change their wrong answer to conform to reality and which will argue that the map is wrong?
This is the real test that creationists fail every day.

garry said...

93 here.

James Cranch said...

I got 89.

The whole business reminds me of that quotation of war-criminal-at-large Donald Rumsfeld: There are known knowns. These are things we know that we know. There are known unknowns. That is to say, there are things that we now know we don’t know. But there are also unknown unknowns. These are things we do not know we don’t know.

jason said...

60

I'm so suck. :|

IvanM said...

As I suspected, you can get a perfect 100 by simply marking every question as 50%! Well, not quite, because of the way they treat a sample size of zero in the 100% column. You'll actually score a 50 if you mark every question as 50%. But if you modify this strategy by changing your answer to two questions (one true and one false) whose answers you're certain of, then you'll score 100.

Steve said...

79

Miranda said...

I took the test three times, to test something out.

A. 50% across the board.
B. 50% across the board, except 100% for #1 and #6. (These were easy questions for me)
C. 50% across the board, except for 0% for #1 and #6.

In order, my score was 51 (give or take), 98, then 2.

With these results, I have great suspicions about the test.

Jon said...

97.

IvanM said...

It's hardly suspicious, Miranda. You can't get a stable estimate of the shape of a curve when most of the sample points are in one spot. The results you describe are exactly what's expected.

It would be nice, though, if they would also give error bars, which should be very large in the examples you gave.

Unknown said...

Bayesian Bouffant, FCD asks:

But does "natural gas" mean what they pump out of the ground, or does it mean what they pipe to my house?


Technically, I believe what's pumped into your house is "City gas" (possibly treated natural gas, coal gas or similar, with added odorants).

It is, however, commonly referred to as "natural gas" or just "gas". So, if one wants to be pedantic, I still believe "natural gas" is mostly odorless.

Corey said...

I received a 62. Being Canadian, I feel that the questions were biased towards Americans (questions on number of Presidents, first words in the American Constitution).

Juhan said...

69, from Italy (yes that of Berlusconi)

Jeff said...

I scored 81. This test goes so nicely with the book I'm reading right now, "On Being Certain" by Robert Burton. Great book... I'll post a review on my blog when I'm done.

crf said...

77

One thing about the test: Do people become more sure of their answers as they "get tired" moving through the latter questions?

jason said...

Another book that deals with this stuff is Philip Tetlock's Expert Political Judgment, which I still need to finish. IIRC, he applies exactly this measure to pundits' forecasts of developments in world affairs.

Pelle said...

There where 50 questions, all different i think.
And there was 50 statements - all looking the same.

The instructions where:
Remember: answer 0% if you are sure that the statement is false.
Answer 100% if you are certain that the statement is true.

So i pressed 100% each time and got a score of 51.
This sure was a silly IQ-test.

I did the test once again.
There where the same questions.
And the same statement "You have not answered this question yet" witch
i now denied by clicking radiobutton
0% 50 times. My RQ score became 17.

Can anyone explain the meaning of this test to me?
Is it simply to check if you can read the instructions?

Jeff said...

Pelle, I guess the test is not balanced for number of true/false answers. Maybe there are more true statements than false.

Cathy Sander said...

86: that was somewhat unusual, given that I know lots of stuff [which is true]...but I think I have a [over]cautious disposition.

Unknown said...

Late to the party but I just discovered this hilariously awesome blog and took this quiz to pass the time.

The RQ score ranges from 0 (low RQ) to 100 (high RQ). Your RQ score is 100.

Some really weird questions on the test, haha. My calibration and identity lines perfectly overlap and are on a diagonal. Pretty cool how they measure it really.

Dan said...

I got a 99. Better than I thought it would be. There were only a few questions I knew with certainty. I wonder if you can get a 100 by answering 50% to all the questions.