Q20 - When people show signs

[haeaznboiyoung]

Debated between C and D and did eventually choose C. At the time, I couldn’t pinpoint why... it was more of a gut feeling and C just sounded "more" correct. But now, I think I see the problem and would just like some confirmation.

D is incorrect because it doesn’t directly attack the conclusion. It attacks the evidence that computer programs can correctly make a diagnosis but does not attack the conclusion. Even though EKG data is alone insufficient to enable either the doctor nor the computer to make a correct diagnosis, the computer still may be preferable because it can do other things with the data better than the doctor (interpret the data more efficiently to contribute to a correct diagnosis for example)

C is correct because it attacks the conclusion directly. It gives additional evidence to show that computers have succeeded over the doctor in only one aspect whereas the doctor succeeds over the computer in the other. Therefore, it weakens the conclusion that computers should interpret and doctors should not.

Yah I probably got the reasoning wrong but it’s all I could come up with lol

[ManhattanPrepLSAT1]

Your reasoning for answer choice (C) is right on. To help others see this as well, let's put some numbers on this.

Suppose 50% of those who had a suspected heart attack actually had a heart attack. Let's also say that in a study of 100 people who had a suspected heart attack the computer diagnosed 48/50 who had a heart attack correctly but also claimed that 42/50 who did not have a heart attack did have one. That would put the accuracy of the computer way off. it's like playing roullette with money on every number. Sure you'll win, but you'll always lose way more than you win.

The doctor may be a bit more cautious with his/her diagnoses but may in fact be more accurate overall. So if the doctor only diagnosed 45/50 who had a heart attack correctly and 49/50 of those who did not have a heart attack correctly, then the doctor would have a much better overall accuracy.

(D) says that EKG data alone are not enough to make an accurate diagnosis, but nowhere does the argument assume that computer programs or doctors make such diagnoses on EKG data alone.

Great work!

[wjun3]

why is (E) wrong?

doesnt it mess up the reasoning because hes saying that computers should be used to intrepret, but we only used one cardiologist in this study?

[wjun3]

okay nevermind i think I got it afte reading the stimulus again... It says "a highly experienced and skilled cardiologist" ...

so would E be correct if it didn't say that?

[ManhattanPrepLSAT1]

Not exactly. One potential weakness in the argument would have been that it compared an unqualified cardiologist against a computer and then concluded that cardiologists in general should not be assigned the task of interpreting EKG data. That would have left open the possibility that a more qualified cardiologist would have done a reasonable job at interpreting EKG data.

However, in this argument they pitted a highly qualified cardiologist against the computer program. Since the cardiologist still did worse (at least with respect to diagnosing a lower percentage of those who later had confirmed heart attacks) than the computer program then it cannot be assumed that a less qualified cardiologist would do any better.

The flaw your going at is whether the sample in the study was representative of cardiologists in general. We know that the cardiologist in the study was not representative of cardiologists in general, but since the cardiologist in the study was more qualified than cardiologists in general, we don't need to worry about the idea that in general cardiologists would do a better job than the cardiologist in the study.

Does that answer your question?

[interestedintacos]

To put this one more simply: a lot of 'confirmed positives' means absolutely nothing if there are a lot of false positives. You need to know both sides of the coin. This sort of reasoning has shown up a number of times in the past.

The last one I remember is a one about a news station that correctly predicted that it would rain more times than other stations, but the info left out how many false positives there were (predicting rain when it didn't rain).

[chike_eze]

why is (E) wrong?

doesn't it mess up the reasoning because hes saying that computers should be used to interpret, but we only used one cardiologist in this study?

This answer choice got me too. I feel like this plays on the Man vs. Machine dynamic. Grrhhh... Man must overcome the machines, Skynet must be destroyed!

I incorrectly thought: "huh, I guess the cardiologist in question wasn't that smart... others may have done better". I totally forgot about the "highly skilled and experienced" qualifier attributed to the cardiologist.

It's uncanny how these LSAT folks come up with these answer choices. It's like they have a wrong choice for every common faulty assumption/bias.

(C) makes sense now. Thanks!

[shirando21]

Your reasoning for answer choice (C) is right on. To help others see this as well, let's put some numbers on this.

Suppose 50% of those who had a suspected heart attack actually had a heart attack. Let's also say that in a study of 100 people who had a suspected heart attack the computer diagnosed 48/50 who had a heart attack correctly but also claimed that 42/50 who did not have a heart attack did have one. That would put the accuracy of the computer way off. it's like playing roullette with money on every number. Sure you'll win, but you'll always lose way more than you win.

The doctor may be a bit more cautious with his/her diagnoses but may in fact be more accurate overall. So if the doctor only diagnosed 45/50 who had a heart attack correctly and 49/50 of those who did not have a heart attack correctly, then the doctor would have a much better overall accuracy.

(D) says that EKG data alone are not enough to make an accurate diagnosis, but nowhere does the argument assume that computer programs or doctors make such diagnoses on EKG data alone.

Great work!

I have no medical background at all.

do you mean that the accuracy of diagnose consists of two parts:

1) the accuracy of the diagnose of the disease

2) the accuracy of the diagnose that it is not this disease (which I did not think it is relevant...)

?

[shirando21]

by choosing D, we are assuming that there are some other methods that can only be used by cardiologists not computers, to help diagnose heart attack.

Is it wrong because we are not supposed to make this assumption?

[ManhattanPrepLSAT1]

do you mean that the accuracy of diagnose consists of two parts:

1) the accuracy of the diagnose of the disease

2) the accuracy of the diagnose that it is not this disease (which I did not think it is relevant...)

The accuracy includes two parts:

1. how accurately do you say the person had a heart attack, when the person actually did have a heart attack?
2. how accurately do you say the person had a heart attack, when the peson actually did NOT have a heart attack?

Suppose the machine always says no matter what that the person had a heart attack. Then every time the person actually had a heart attack, the machine would be correct. But it would also be incorrect every time the person had not actually had a heart attack!

by choosing D, we are assuming that there are some other methods that can only be used by cardiologists not computers, to help diagnose heart attack.

Is it wrong because we are not supposed to make this assumption?

Answer choice (D) does not actually make this assumption. Take it for what it says - for some cases, EKG data alone is not enough for either doctors or machines to make accurate diagnoses. That's consistent with the original argument, since it never claimed (nor assumed) that either machines or doctors were 100% accurate.

[oa246]

With E, it should be noted that the argument explicitly tells us that the cardiologist is highly skilled. So, if E is true and this cardiologist is unrepresentative of cardiologists in general, then that would strengthen the conclusion that computer programs should interpret EKG data. After all, if this highly skilled cardiologist is an anomaly, how well would the other cardiologists interpret the data vs the machine?

[asmaa737]

Why is (A) wrong?

Here was my reasoning:
"Obvious" means those mistakes usually won't be repeated. Maybe he was really tired that day and his incorrect diagnosis was not representative of how he was that day?

The only reasons I can think of as to why (A) might be wrong is that "few" could mean just 1 mistake...which would strengthen the argument b/c then it shows that he only made one mistake and was still not as accurate as the computer. Also, this doesn't say anything about what big mistakes he made.

[christine.defenbaugh]

Interesting point you raise, asmaa737!

There are two serious problems in your analysis of (A) though. First you're putting a lot of information into the word "obvious". You have to make a lot of assumptions to conclude that obvious mistakes are ones that won't be repeated, or ones the cardiologist wouldn't make on a normal day. There's nothing given to us here that supports that.

But even if that were supported, (A) emphasizes that the cardiologist "made few" of these obvious mistakes. It's not just that it might have been 1 mistake, but rather this statement underscores that the number of obvious mistakes was notably small - and with your definition of 'obvious', that would strengthen argument, not weaken it!

Does that make sense?





There have been a ton of great thoughts, and great questions in this thread, and I'd like to consolidate everything into one complete explanation, for the sake of future readers!

As with any assumption family question, you should attack this weaken question by first distilling the core of the argument.

PREMISES
A computer program correctly diagnosed more confirmed heart attacks than did a highly skilled cardiologist.
CONCLUSION
Computer programs should be used to diagnose stead of cardiologists.

The disconnect arises between the correct diagnosis of actual heart attacks and 'so they should always be used'. In other words, the argument assumes that there's no other possible consideration that could weigh against the idea of using the computer program. To that end, (C) brings up a pretty significant concern!

The correct diagnosis of the NON-heart attacks matters too! If (C) is true, then the cardiologist might have a much better error rate overall! What if there were 100 confirmed heart attacks, and 100 NON-heart attacks? The situation might look like this:

----------------------Heart Attacks ----- NON Heart Attacks
Computer correct--------100---------------------0
Cardiologist correct-------90--------------------90

The computer would only have a 50% accuracy rating overall, while the cardiologist might have a 90% accuracy rating overall! That would seriously call into question the conclusion that it should all be left to the machines!


Non-Weakeners
(A) The obviousness of any mistakes made by the cardiologist is not relevant to conclusion. Don't invest more meaning into 'obvious' than appropriate!
(B) There's no evidence that subjective judgments have any place in interpreting EKG data. The conclusion is not suggesting that computers should take over the entire practice of medicine, just EKG interpretation!
(D) This increases the error or failure rate, but does so for both man and machine. Since the conclusion is about preferring the machine to man, this doesn't affect that comparison.
(E) We already know he was unrepresentative - he's 'highly skilled'! If he's better than your average Joe cardiologist, and he still couldn't beat the machine, that strengthens the argument that we should leave this job to the computer!


I hope this helps!

[aznriceboi17]

Do you think the LSAT authors considered the severity of errors in heart-attack vs no heart attack cases?

When I read this question, my thought was that of the two error types, a false-positive (saying there's a heart attack when there isn't) is less severe than a false-negative (saying there's no heart attack when there isn't). My reasoning for this was that doing nothing while a heart attack is going on is obviously bad, while using heart-attack treatment methods on a person when there is in fact no heart-attack is NOT obviously bad (now that I write this, it seems less plausible... ). If C were true, the fact that the cardiologist had a LOWER false-positive rate than the computer means that it had a HIGHER false-negative rate, which would suggest that the cardiologist would be a bad option since false-negatives are worse than false-positives, thus strengthening the argument. This is why I didn't pick it.

I was also drawn to D because it seemed to say that there are a lot of cases where EKG data isn't enough, you need to look at other signals as well, such as physical symptoms. People can readily take advantage of these additional signals, but modifying the computer program to do so would be much more difficult. This would suggest that cardiologists are still preferable for interpreting EKG data since in many cases it will be evaluated in a context where other signals are required.

Any general advice for avoiding this kind of flawed thinking?

[christine.defenbaugh]

Thanks so much for posting, aznriceboi17!

You raise some very worthwhile concerns here. What's most interesting about them is that you managed to talk yourself out of choosing (C), and into choosing (D) for the same fundamental reason: you were adding too much information to the answer choice in each case.

The LSAT writers could absolutely have addressed the issue of the difference in severity between false-positives and false-negatives. But since they didn't as of yet, it's unwise for us to go about making assumptions about it. I think you started to realize this as you were writing it out - you did a lot of mental work to justify that one error was more significant than the other, but don't actually know if any of that holds true!

But let's live in a world for a moment where false-negatives are REALLY BAD and false-positives are no big deal. The premises already tell us that the physician has a higher false-negative rate than the computer program. So, that's pretty bad. Answer choice (C) just tells us that the physician has a lower false-positive rate. If false-positives are no big deal, then this information doesn't strengthen the argument - it would do nothing at all.

The point here is that there's no way this information will strengthen the argument - if false-positives are bad it weakens the argument, and if they are no big deal, the argument is left untouched. Since we don't know which one of these situations is actually true without making some assumptions, we have to accept that both are possible - and that means that this answer makes the conclusion at least a bit less likely!

I think you made two essential errors in assessing (C): you made assumptions about the relative damage of false-positives vs false-negatives, and you thought the answer choice lead you to the information that the physician has a higher false-negative rate - we already had that information from the premise!

Interestingly, you also added in some assumptions of your own in assessing (D)!

...you need to look at other signals as well, such as physical symptoms. People can readily take advantage of these additional signals, but modifying the computer program to do so would be much more difficult.

Those 'other signals' that you might need to look at, how do you know that people would be better at reading those signals than computer programs? Couldn't the additional signals be something that a human couldn't detect, but a computer program could? Say, a change in body temperature by a fraction of a degree? If that were the case, this answer would strengthen the argument, not weaken it!

Since the answer could strengthen or weaken the argument depending on which assumptions we make, we can't say that on average is makes the conclusion any more or less likely! There's just no way to know!

The good news here for you, as you continue your LSAT journey, is that the culprit on both of these analyses is the same: you're adding in your own biases/outside information/assumptions into your assessment of the answer choices.

The only cure for this is to start becoming more sensitive to when you are making an assumption. Challenge yourself to provide support for any reasoning that you are making. If you identify an assumption that you are making, don't just kill it with fire - instead, examine that assumption: what if the assumption you are making wasn't true? Does that drastically change how the answer choice affects the argument?

Does that help clear things up a bit?

[aznriceboi17]

Yes that was very helpful, thanks Christine!

Also, thank you for reading through my post and catching the error I made regarding C and the lower false-positive rate of the physician -- I think I lost track of the terms and got confused.

[zhangjiwen325]

[quote="mattsherman"]Your reasoning for answer choice (C) is right on. To help others see this as well, let's put some numbers on this.

Suppose 50% of those who had a suspected heart attack actually had a heart attack. Let's also say that in a study of 100 people who had a suspected heart attack the computer diagnosed 48/50 who had a heart attack correctly but also claimed that 42/50 who did not have a heart attack did have one. That would put the accuracy of the computer way off. it's like playing roullette with money on every number. Sure you'll win, but you'll always lose way more than you win.

The doctor may be a bit more cautious with his/her diagnoses but may in fact be more accurate overall. So if the doctor only diagnosed 45/50 who had a heart attack correctly and 49/50 of those who did not have a heart attack correctly, then the doctor would have a much better overall accuracy.



I do have a follow up Q regarding answer choice (C).

The conclusion in the stem is "Interpreting EKG data, therefore, should be left to computer programs." So the point here is to interpret data, but not trying to attain a higher overall accuracy of heart attack. Moreover, it seems to me that EKG focuses on people who show signs of having a heart attack, not those who have not shown any signs. Thus (C), which indicates that cardiologist does a better job in cases which no heart attack occured than did computer program just seems irrelevant.

Could anyone correct and answer me pls? Thanks in advance.

[g1oriaaa]

I think there were scattered responses that hit the nail on the head.
The conclusion is that interpreting EKG data should be left to computer programs. The support given is that the program correctly diagnosed a significantly higher proportion of the cases that were later conformed to be heart attacks than the highly experienced doctor did. Our task is to weaken.

A. is wrong because so what? This like saying give me a 180 on the LSAT even though I made a few obvious mistakes that the experts can point out while the computer scored a 180. Clearly it did better than me (hopefully relatable to everyone reading this post).
B. Whether this is true or not does not have to do with the specific topic in question which is interpreting EKG data. This is not a subjective judgement. This is objective. (or as objective doctors claim their field of study is...but that is a whole different philosophical debate)
C. Yay for right answer! This is basically leaves room for the fact that maybe it was the case that the computer just yes to all cases so of course the proportion could be higher. The doctor maybe got ONE wrong by saying it was not a heart attack when it was. But the computer said yes to all so it had a higher proportion of guessing correctly that it was a heart attack but on that same note all those that were not a heart attack the computer said it was and that shows that the computer program therefore should not be left to interpret the EKG data.
D. This just weakens the conclusion by saying no EKG data. But that is not the argument. The argument is the connection between who can interpret better. THis is just throwing the whole thing in the trash. Our job on weakening and strengthening questions is not to weaken/strengthen or contradict a premise or conclusion directly but to weaken or strengthen the relationship between the premises and conclusion (or on harder questions premises and subconclusion or subconclusion and conclusion)
E. This was the one that almost got me. It took me a while to realize that this answer choice actually strengthen the stimulus. If this cardiologist is unrepresentative of the cardiologist in general with respect to skill and experience then we are all effed. Maybe if this was a cardiologist that was a drunken doctor who cheated on his MCAT and tests then yes this perhaps could have a potential for a correct answer but the cardiologist in question is noted as being "experience and highly skilled" so if you are saying this guy isn't representative then you are basically saying "yes we need this EKG data to be interpreted by computers" because even one of the best doctors cannot interpret this.