I got the 1982 University of British Columbia ordering right easily, though that might be because I’m already aware of the phenomenon being studied.
It would be much harder for me as a subject to deal properly with the Second International Congress on Forecasting experiment. Even if I’m aware that adding or removing detail can lead my estimate of probability to change in an illogical way, my ability to correct for this is limited. For one thing, it is probably hard to correctly estimate the probability that I would have assigned to a more (or less) detailed scenario. So I may just have the one probability estimate available to me to work with. If I tell myself, “I would have assigned a lower probability to a less detailed scenario”, that by itself does not tell me how much lower, so it doesn’t really help me to decide whether and how much I should adjust my probability estimate to correct for this. Furthermore, even if I were somehow able to accurately estimate the probabilities I would have assigned to scenarios with varying levels of detail, that still would not tell me what probability I should assign. If my high-detail assigned probability is illogically higher than the low-detail assigned probability, that doesn’t tell me whether it is the low-detail probability that is off, or the high-detail probability that is off.
So as someone trying to correct for the “conjunction fallacy” in a situation like that of the Congress in Forecasting experiment, I’m still pretty helpless.
I got the 1982 University of British Columbia ordering right easily, though that might be because I’m already aware of the phenomenon being studied.
It would be much harder for me as a subject to deal properly with the Second International Congress on Forecasting experiment. Even if I’m aware that adding or removing detail can lead my estimate of probability to change in an illogical way, my ability to correct for this is limited. For one thing, it is probably hard to correctly estimate the probability that I would have assigned to a more (or less) detailed scenario. So I may just have the one probability estimate available to me to work with. If I tell myself, “I would have assigned a lower probability to a less detailed scenario”, that by itself does not tell me how much lower, so it doesn’t really help me to decide whether and how much I should adjust my probability estimate to correct for this. Furthermore, even if I were somehow able to accurately estimate the probabilities I would have assigned to scenarios with varying levels of detail, that still would not tell me what probability I should assign. If my high-detail assigned probability is illogically higher than the low-detail assigned probability, that doesn’t tell me whether it is the low-detail probability that is off, or the high-detail probability that is off.
So as someone trying to correct for the “conjunction fallacy” in a situation like that of the Congress in Forecasting experiment, I’m still pretty helpless.