Is this really a fallacy? In the USSR and Poland case, we might take the probability space in (1) to exclude an invasion of Poland, and the space in (2) to include one. Then the claims are perfectly consistent, since the probability space changes; people just reason with respect to “stereotypical” alternatives.
we might take the probability space in (1) to exclude an invasion of Poland, and the space in (2) to include one
That seems like an unjustified interpretation, since, according to the OP:
Two different experimental groups were respectively asked to rate the probability of two different statements, each group seeing only one statement:
“A complete suspension of diplomatic relations between the USA and the Soviet Union, sometime in 1983.”
“A Russian invasion of Poland, and a complete suspension of diplomatic relations between the USA and the Soviet Union, sometime in 1983.”
Since the subjects receiving statement 1 do not even see statement 2, they would have no reason to exclude the possibility of an invasion of Poland from statement 1.
I believe this is not a conjunction fallacy, but for a different reason. In the first case, the test subject is required to conceive of a reason that might lead to a complete suspension of relations. There are many different choices, invasion, provocation, oil embargo of Europe, etc. Each of these seems remote, that the test subject might not even contemplate them. In the second case, the test is given a more specific, therefore more conceivable sequence of events.
A good third scenario, to control for this, would have been to ask another group of subjects the probabilities independently: A. That USSR invades Poland
B. That US suspends relations
This provide the same trigger of a plausible provocation, but doesn’t directly link them. Variances between the estimates of B in this case v. 1 in the original test would indicate confidence interval between variances between 1 and 2.
Is this really a fallacy? In the USSR and Poland case, we might take the probability space in (1) to exclude an invasion of Poland, and the space in (2) to include one. Then the claims are perfectly consistent, since the probability space changes; people just reason with respect to “stereotypical” alternatives.
Okay, Eliezer should add a boldfaced note at the bottom of this post asking people not to comment until they’ve read the followup.
The answers provided there seem not to be relevant to the objection I raised.
That seems like an unjustified interpretation, since, according to the OP:
Since the subjects receiving statement 1 do not even see statement 2, they would have no reason to exclude the possibility of an invasion of Poland from statement 1.
I believe this is not a conjunction fallacy, but for a different reason. In the first case, the test subject is required to conceive of a reason that might lead to a complete suspension of relations. There are many different choices, invasion, provocation, oil embargo of Europe, etc. Each of these seems remote, that the test subject might not even contemplate them. In the second case, the test is given a more specific, therefore more conceivable sequence of events.
A good third scenario, to control for this, would have been to ask another group of subjects the probabilities independently:
A. That USSR invades Poland B. That US suspends relations
This provide the same trigger of a plausible provocation, but doesn’t directly link them. Variances between the estimates of B in this case v. 1 in the original test would indicate confidence interval between variances between 1 and 2.