Hacker News Reader

Comment by morpheuskafka | Hacker News Reader

Unless its some kind of trick question, shouldn't the first question be asking which is the most detailed, not the must accurate? If all of them are completely in accordance with the given true information, then they are all equally and perfectly accurate, even if one just says "the sky was blue."

View on HN

Yeah I was a bit confused by it. The only way you can judge accuracy here is either from the base source, Jose, or the between inter-witness congruity. But accuracy is bad terminology for this. Given another prompt:

Officers Joe Schmoe and Tits McGee arrive at the scene where Josb has a bloody face and says that the sky suddenly turned red and there are dolphins flying around. There are three witnesses. Witness 1 says there are pink and green dolphins flying around and attacked the man. Witness 2 says the sky turned green and two green dolphins attacked the man. Witness 3 says the sky turned pink and two pink dolphins attacked the man. No witness mentions the bag of mushrooms on the ground. Which account is the least accurate.

According to the information above, what is Josb's brother's name?

This reminds me of the DMV test when I moved to Oregon. The question was "You see a motorcycle coming towards you and its left blinker is on. Should you assume a) it is turning left, b) it is turning right, c) it is not turning, or d) none of the above." The correct answer was c with the explanation that motorcycle turn signals frequently don't automatically turn off. Despite this being literally the most dangerous answer.

If a vehicle is approaching you head on, isn’t it much safer to assume they aren’t turning? You’ll usually be wrong, but if you are right it could avoid a head on collision

It its coming towards you in the other lane, which is the most likely circumstances for it to be moving towards you … Ignoring a turn signal sounds like a great way to have a head on collision.

I read it as, it's coming towards you in the other lane and has to cross your lane to turn left. If you "assume" it's not turning, you can just keep your speed rather than maybe slowing down just in case he doesn't see you coming.

> just in case he doesn't see you coming.

Which results in a dead motorcyclist.

It is presumably in the other lane. But let's look at modes of failure.

- Assuming the light is properly indicated but in this failure mode the motorcyclist turns. You collide and the motorcyclist faces serious injuries or dies.

- Assuming the light isn't properly indicating intention, you slow down, and cars behind you might also slow down. No one injured.

There is clearly one option that is significantly safer, even if the probability is low.

You're the type of person it's trying to weed out - you're too smart.

https://abcnews.go.com/US/court-oks-barring-high-iqs-cops/st...

Analysis paralysis is not an indication of smart.

Pointing out "the question is wrong/invalid/badly worded" isn't analysis paralysis

Could be, of course, that they want to test how you deal with that... But I have a feeling that's beyond the point here

Analysis paralysis is evidence of somebody either completely missing the point or people who wish to think they are smarter than they can actually perform (Dunning-Kruger). Either way it is a performance failure, and failing to perform against a baseline, even if flawed, is not an indication of smart.

Good thing it wasn't analysis paralysis. The conclusion of the analysis was "the question is badly worded" and from the way he worded his original reply it seems like he understood what they were aiming for

That is an (small) indication of smart

Sounds to me like just an oversight. If what experience I have with public institutions is worth anything, it just didn't occur to the people who designed the question that the wording was shit and that to arrive at the intended answer you have to be loose with the meaning of words. That does not reflect well on American po-pos

why do you keep posting this over and over again in this thread? The department applied this rule to attempt to reduce turnover, not because they literally want dumb cops.

I never said they want dumb cops, just that the person posting is not who they want because they're too smart. The way they limit turnover is by selecting for normal IQs so they don't get bored and leave.

Seems like the smart thing to do would be to give the "too smart" people training to tackle more interesting and important problems - like white collar crime, tax loopholes, internal corruption, etc.

I'll give you three guesses why that's never even floated as an option...

To be fair, the white collar crime is mostly tackled by other organizations who have investigators but aren't typical police. Like the FBI (accountants comprise a large percentage of agents), SEC, IRS, etc.

>> shouldn't the first question be asking which is the most detailed, not the must accurate?

The first question askes which witness is the LEAST accurate.

One statement is at odds with others in many ways, ergo:

* that statement is least accurate, OR

* the victim and two witnesses are colluding on a false narrative to cover the actual events because ... ??? ( poss. known assailant, fear of retailation, witnesses are part of gang pressuring victim, etc. )

Witness 2 gives detailed information, but it's completely unlike the information given by both the victim and the other 2 witnesses. The victim, Witness 1 and 3 present the same story, although Witness 3's is very sparse on details (ie least detailed, according to your metric).

You may also scroll down to the end of the PDF to check the answer. Witness 2 is the least accurate.

I get it. But without knowing what actually happened it's impossible to know who was the least accurate, there's nothing stopping Jose, W1, and W3 being mistaken.

I feel it really should say "likely to be the least accurate".

No, Witness 2 has known details wrong. We can assume the officers are standing on West Street with the victim. Unless Witness 2 says that the victim moved from West street to Canal St, Witness 2 is the least accurate. Accuracy has nothing to do with the description of the assailant, as that can't be held to any level of accuracy.

“To a cop, the explanation is never that complicated. It's always simple. There's no mystery on the street, no arch-criminal behind it all. If you got a dead guy and you think his brother did it, you're going to find out you're right.”

Writing good tests hard. Really hard. Unfortunately most people think "Bill knows about XYZ, he can write the test."

The results are frequently not good.

Reply on news.ycombinator.com