QUESTION TYPE: Weaken
CONCLUSION: Instead of using cardiologists, we should use computer programs to interpret EKG data.
REASONING: A computer program identified a higher percentage of heart attacks than a cardiologist did.
ANALYSIS: There are two sides to identifying heart attacks:
- Correctly saying that a situation is a heart attack.
- Correctly saying that a situation is not a heart attack.
Though this stimulus is worded to sound convincing, the setup only tells us about the first bullet point. We don’t know how the computer performed at identifying situations that weren’t heart attacks.
Suppose you had 1000 people, and 45 were having heart attacks. Ideally, you’d want to label the 45 people as “having heart attacks” and the other 955 people as “not having heart attacks”.
A cardiologist might catch 43 of the heart attacks, miss two, and incorrectly say that 3 of the non-heart attack patients had heart attacks. Not bad, overall.
What do we know about the computer program? All we know is that of the 45 people with heart attacks, it correctly labelled more of them. So, 44 or 45. A bit better. But we don’t know how the computer did with the other 955 people. If the program also said that 900 of the healthy people were having a heart attack, then it would be a useless program. It thinks everyone is having a heart attack!
A good program or cardiologist must identify both when a heart attack is happening and when one is not happening.
- So? The cardiologist could be good, but if the computer were better then we should still use the computer.
- This is a tempting answer. But the first half is fluff. The second half says computers aren’t good at subjective judgements….but is interpreting EKG data a subjective judgement? If EKG interpretations are objective judgements, then there’s no reason not to use computers.
- CORRECT. This destroys the argument. This shows that the computer had a high “false positive” rate. I.e. The computer was sending healthy people to the hospital, while the cardiologist correctly determined that many people were not having heart attacks.
You could make a computer program that says every situation was a heart attack. It would identify 100% of heart attacks, but still be a useless program because it would have an atrocious false positive rate.
- This answer tells us that some situations are impossible to judge. Who cares? The question is only about whether to use a cardiologist or a computer in the situations where EKG data do allow us to make a decision.
- Another trap answer. This plays on your mistaken assumption that studies should always use representative examples. This isn’t true.
In this case, the cardiologist was “very experience, highly skilled”. And it seems a computer still beat them. So this answer strengthens the conclusion. If even an excellent cardiologist can’t beat a computer, then what hope does a normal cardiologist have?
Want advanced LR strategies? → Try the LR Mastery Seminar
Graeme shows how 170+ scorers do LR in real time. Satisfaction guaranteed!