> While on a drug, participants made a poorer, more random first attempt at filling a knapsack than they did after taking the placebo, and this had an especially negative impact on the subsequent performance of those who’d been above average in the placebo condition, explaining why they dropped below the mean. Overall, the results suggest that the participants’ approach to solving the knapsack task became less systematic while they were on each of the three drugs, the researchers write.
I can't help but think that since the knapsack problem is NP-complete, any attempt to fill the sack could subjectively appear "more random." Certainly the participants could use certain heuristics, but since in general there is no known way to systematically solve the problem, this study seems to enable some confirmation bias on the part of the experimenters.
NP-complete just means that it’s very difficult to verify that any solution is the best possible solution. It’s very easy to compare any two attempts at a solution. That’s what the researchers did, they saw that the knapsacks from the drugged attempts were worse than placebo
That’s an interesting observation. However, maybe that makes it the perfect type of problem. That is, it’s a problem you can spend a lot more time on without better results, on average.
I can't help but think that since the knapsack problem is NP-complete, any attempt to fill the sack could subjectively appear "more random." Certainly the participants could use certain heuristics, but since in general there is no known way to systematically solve the problem, this study seems to enable some confirmation bias on the part of the experimenters.