Well explained and very recognizable!
This is so helpful!
Very interesting - I can relate to it. I noticed a typo in the blog ( minor one) - "anwser" ( instead of answer)
@kinglovesqueens said in #4:
> Very interesting - I can relate to it. I noticed a typo in the blog ( minor one) - "anwser" ( instead of answer)
Thank you! I make that mistake again and again (I actually did it twice just on this post :D)
> Very interesting - I can relate to it. I noticed a typo in the blog ( minor one) - "anwser" ( instead of answer)
Thank you! I make that mistake again and again (I actually did it twice just on this post :D)
One thing that I'd add is that 2-4-6 experiment in the blog post isn't the same as the original as in the original only 1 sequence is given while in the blog post 4 are given each with the increment by 2 criteria that does mean misdirection is stronger in your case and that could be down to Asch conformity experiments with simulations. This does mean people are less likely to try 3-6-9 assuming it's a ratio.
Actually way I'd look at it you would need to give literally 1000s of 3 number combinations to be relatively sure of the exact rule. This is because it's entirely possible that numbers in ascending order isn't the only rule when it could follow AND or OR conditions. It could instead be something like numbers in ascending order and at least one number is non-prime.
I think the idea of the experiment is that people assume it's increment + 2 not realising it's a subset of the ascending order rule. However there's nothing stopping the ascending order rule being a subset of another rule. So then question is at what point do you stop searching for a rule that may not exist. So it can be argued that everyone that found the "correct" rule also has succumbed to confirmation bias.
However in example given one thing missing was a case where pattern didn't apply so if you don't have a failure condition anything could work. If exhaustively testing things to consider include numbers within a set range, there's no clarification as to how numbers that aren't integers are treated (e.g. 0.5, 1/3, pi, infinity, imaginary numbers etc), whether negatives are treated as absolute values or not etc.
Actually way I'd look at it you would need to give literally 1000s of 3 number combinations to be relatively sure of the exact rule. This is because it's entirely possible that numbers in ascending order isn't the only rule when it could follow AND or OR conditions. It could instead be something like numbers in ascending order and at least one number is non-prime.
I think the idea of the experiment is that people assume it's increment + 2 not realising it's a subset of the ascending order rule. However there's nothing stopping the ascending order rule being a subset of another rule. So then question is at what point do you stop searching for a rule that may not exist. So it can be argued that everyone that found the "correct" rule also has succumbed to confirmation bias.
However in example given one thing missing was a case where pattern didn't apply so if you don't have a failure condition anything could work. If exhaustively testing things to consider include numbers within a set range, there's no clarification as to how numbers that aren't integers are treated (e.g. 0.5, 1/3, pi, infinity, imaginary numbers etc), whether negatives are treated as absolute values or not etc.
rook into d8 mate
queen h7 mate
@KMcGeoch said in #6:
>
The experiment could never be the same, as I do not actually have a way to get feedback from you. As mentioned in the post, I would instead " simulate" your replies - of course, if you are already aware of the possibility of falling into confirmation bias, you will most likely spot that my replies for you are not trying to refute an initial hypothesis.
Afterwards, I indicate the original study and mentioned how the initial sequence was simply "2-4-6" and test subjects typically did not try to refute their initial hypothesis.
As for your next point, I just do not agree. Sure, you need to give thousands of hypotheses in order to be *sure*, but in order to be reasonably sure, not really. I am not sure if your example of " ascending order and one non-prime number" would be allowed in the study, as that seems to me like two rules instead of one.
However, I do not think that this is an important point. The point of that section of the post is to play a little game with the reader that might or might not engage them with the possibility of Confirmation Bias, and then introduce them to the original experiment where that ideas was "verified". I am more concerned with the implications this has in chess, than if I managed to "win" the game I played with the reader :).
>
The experiment could never be the same, as I do not actually have a way to get feedback from you. As mentioned in the post, I would instead " simulate" your replies - of course, if you are already aware of the possibility of falling into confirmation bias, you will most likely spot that my replies for you are not trying to refute an initial hypothesis.
Afterwards, I indicate the original study and mentioned how the initial sequence was simply "2-4-6" and test subjects typically did not try to refute their initial hypothesis.
As for your next point, I just do not agree. Sure, you need to give thousands of hypotheses in order to be *sure*, but in order to be reasonably sure, not really. I am not sure if your example of " ascending order and one non-prime number" would be allowed in the study, as that seems to me like two rules instead of one.
However, I do not think that this is an important point. The point of that section of the post is to play a little game with the reader that might or might not engage them with the possibility of Confirmation Bias, and then introduce them to the original experiment where that ideas was "verified". I am more concerned with the implications this has in chess, than if I managed to "win" the game I played with the reader :).
@HelloItsDmitri said in #9:
> As for your next point, I just do not agree. Sure, you need to give thousands of hypotheses in order to be *sure*, but in order to be reasonably sure, not really. I am not sure if your example of " ascending order and one non-prime number" would be allowed in the study, as that seems to me like two rules instead of one.
Well it really depends on what you define as a rule and what is specified at the start of the study. From my understanding psychological approach of the study is that when it's not defined that people make assumptions and when they see results that match these assumptions they take it as confirmation. These assumptions include things like "There's a formula to calculate next value" and "There's only 1 clause in the rule". Moment you specify in depth exactly what relations are psychological approach fails since you provide a checklist to check against. It's not unusual to have multiple clauses in a rule with for example rule for getting the WGM title having a sex, norms and rating requirement.
If you want it converted into chess implications basically whenever anyone plays a move normally they don't calculate all the variations until checkmate or a draw. Instead humans will evaluate from a critical to a settled position that two different players could evaluate differently. It's also why engines/tablebase beats humans as it calculates further so will check all the conditions a human doesn't and is also the reason that after games humans use engines to verify that their assumptions were correct. It's also not uncommon for players to spot one trap and assuming they've avoided it they play something else that falls into a different trap
Probably main implication comes into consideration of candidate moves where often a strong move isn't played because it wasn't considered. This most frequently occurs for piece sacrifices and missed tactics. So when I fail to play the best move questions I ask myself is either why didn't I consider it or alternatively why didn't I play it when I did consider it.
> As for your next point, I just do not agree. Sure, you need to give thousands of hypotheses in order to be *sure*, but in order to be reasonably sure, not really. I am not sure if your example of " ascending order and one non-prime number" would be allowed in the study, as that seems to me like two rules instead of one.
Well it really depends on what you define as a rule and what is specified at the start of the study. From my understanding psychological approach of the study is that when it's not defined that people make assumptions and when they see results that match these assumptions they take it as confirmation. These assumptions include things like "There's a formula to calculate next value" and "There's only 1 clause in the rule". Moment you specify in depth exactly what relations are psychological approach fails since you provide a checklist to check against. It's not unusual to have multiple clauses in a rule with for example rule for getting the WGM title having a sex, norms and rating requirement.
If you want it converted into chess implications basically whenever anyone plays a move normally they don't calculate all the variations until checkmate or a draw. Instead humans will evaluate from a critical to a settled position that two different players could evaluate differently. It's also why engines/tablebase beats humans as it calculates further so will check all the conditions a human doesn't and is also the reason that after games humans use engines to verify that their assumptions were correct. It's also not uncommon for players to spot one trap and assuming they've avoided it they play something else that falls into a different trap
Probably main implication comes into consideration of candidate moves where often a strong move isn't played because it wasn't considered. This most frequently occurs for piece sacrifices and missed tactics. So when I fail to play the best move questions I ask myself is either why didn't I consider it or alternatively why didn't I play it when I did consider it.