Steve Spencer
Offline
• • • • • • •
Upload & Sell: On
|
 p.2 #14 · fuji medium format compared to others | |
I know exactly what confirmation bias is and how difficult it is to fight. We all contend with it to some extent, and that’s why things like peer review are critical in the academic world, for example.
But holding a position is not the same as exhibiting confirmation bias.
In fact, the experiment I did with print files from FF and miniMF systems several years back was designed specifically to avoid confirmation bias. It was what is sometimes called an ABX test, in which observers who are not told the objective of the test nor the sources of the samples to compare are asked to compare three samples A, B, and X. A and B come from different sources, and X could come from either A or B but which one is not indicated.
Observers are asked to note differences they observe between A and B and to express a preference for one or the other. Then they are asked whether sample X is from the same source as either A or B. As you can probably anticipate, their ability to correctly answer that second question tends to cancel out potential confirmation bias in their answer to the first A/B question.
In my test, participants were provided with three image samples — in some cases in the form of prints, in. other (online) cases in the form of print files that they could print themselves. The samples came from a 50MP 5DsR and a 50MP GFX50S, both with roughly equivalent native lenses. The test images were shot with the systems on tripods and with all available procedures to ensure stability and accurate focus.
Then the images were prepared for printing in usual ways, at a range of sizes from (IIRC) 12 x 18 (or a f 4:3 equivalent, I’d have to look it up again) up to 40” x 60”. (For my own edification I actually did a comparison at 80" x 120", though that was not part of the public test.) Then produced equal sized crops from each of the full size images, in sizes that could be printed on letter-sized stock.
I was interested in seeing at what print size observers, as a group, could reliably express a preference for one source (A or B) over the other, with reliability determined in part by their ability to accurately determine whether X matched A or B. In the end, if memory serves, it wasn’t until we got to the 40” x 60” sourced images that the percentage choosing the sample from the miniMF system began to be higher than random probability.
So, yes, I understand what confirmation bias is, how susceptible we all (including you) are to its effects, and how we can try to see past it. (Speaking of confirmation bias, I note that only fevel that clam against those who hold a different point of view than yours. That suggests... confirmation bias! ;-) )
And, again, your point about this bias was also essentially irrelevant to the post your replied to.
I had a funny revelation the other day while thinking about some of my interactions with you on FM. I am, admittedly, a bit (I can see the eye rolls) a “bit” pedantic about some things here. And, yes, my posts can be pretty long and may require more than a cursory scan to understand.
The revelation was that you might be one of the “select group”here who can compete with and even exceed on occasion my level of what some will describe as pedantry… ;-)
Congratulations....Show more →
Dan, I mean no ill will in pointing out the confirmation bias in your post. Am I pedantic? Guilty as charged. Perhaps not required of professors, but many of us are. Do I exhibit confirmation bias? You betcha. I don't just think we all do to some extent, but I think one of the fundamental lessons of my field is that all people are very prone to confirmation bias. Do I regularly accuse people of confirmation bias who disagree with me? That one I don't think I do. In fact, I am pretty sure that you are the only person to whom I have pointed out their confirmation bias and I do that because I know you understand what it is. Even with you, however, I don't think I have not done so for a good 5 years.
Odd that you consider your experiment designed to avoid confirmation bias, because I think your experiment confirmed exactly what you have argued all along: that people can't see differences between the sensor size used. I would argue the design of the study made that result likely to be observed and it doesn't avoid confirmation bias, but instead is a different classic example of confirmation bias. I'll explain if you want, but will avoid all the details to try to cut down on being pedantic. Send me a PM if you want to discuss it off line from this thread. For now let me simply say I don't think that study comes close to showing what you think it shows.
I should hasten to point out, however, that although we have gone back and forth over issues including GF cameras over the years. I have enjoyed interacting with you. I think we agree far far more than we disagree and I hope and I have found that when we disagree we can do so in a civil and friendly fashion. I think as a society we need to be better about both allowing disagreement and disagreeing with each other in such a civil and friendly way. I hope you have understood our interactions in that way too. Perhaps, I have not been as civil as I hoped, however, and for that I apologize.
|
)