View attachment 526082
If you're stating the truth then I'm sure you wouldn't have a problem explaining it in depth rather than just spouting claims without evidence.
Tell me which one of them looks better or dominant and why exactly?
View attachment 525873 View attachment 525875
I know what facial averageness is, I am not an idiot. What I am saying is that these are not actual real averages of an entire population of a certain ethnicity. It's the basic rule in sampling and averageness that when the sample size is too small for a population the results will be skewd and inflated/deflated. In order for an average of a population to be accurate, the sample size used in a study must be big enough to represent the entire population.
There's a difference between "average faces
from a certain population/area" and "average faces
of a certain population/area"
These overlayed pictures (that look like AI btw) are not the average faces of a population, they are average faces
from a population.
Unless there's a way to find out how many overlayed faces were used (how big is the sample size), we can't take these so-called "average faces" for granted.
Also selection bias and margin of error play a role in skewing the results of a population average, selection bias is when you pick specific individuals to be included in your sample instead of doing a random selection. The more random the more accurate, the more specific/cherrypicked the less accurate. Margin of error is the amount of error in the results of a study, the higher the margin of error the less credible/accurate the result is. Since there's also no way we can find out the percentage of selection bias and margin of error in this facial averageness study unless we see the sampling size used in the study then we can't take these average faces for granted.
Now what's the difference between "average faces from a certain population/area" and "average faces
of a certain population/area"? The difference is that you can make an average face out of 4 or 10 or 1000 faces only, while it's technically an average, a sample size of 4 or 10 or 100 or 1000 is not enough to represent an entire population of 1.38 billion currys though.
View attachment 526024
Here's a chart to make things clear, since I don't have time to do complex maths for a population of 1.3 billion right now, just analyzing this chart and reaching a solid conclusion from it should be enough to get the point across.
The maximum population in this sampling table is 1 million so this is what we will focus on since it's the closest to 1.3 billion out of them all,
So basically the ideal sample size for the most accurate result in a 1 million population is 6 different variables ( 384, 1534, 9512, 663 2647, 16317) depending on something called confidence level and margin of error. As you can see the lower the margin of error the higher the sampling size should be (the bigger the sampling size the more credible and accurate as I stated above, that's why margin of error is low when a big sample size is used) and also the bigger the sampling size the higher the confidence level should be. According to dictionary, confidence level is a measure of the reliability of a result. A confidence level of 99 per cent or 0.99 means that there is a probability of at least 99 per cent that the result is reliable.
So basically the ideal sampling size (The amount of individuals being studied from a certain population) for a 1 million population is 16,317 with confidence level of 99% and margin of error of 1%, this is the most accurate you can get. Now this is when the population is 1 million, imagine when the population is 1.3 billion (as it should be). In order to meet a margin of error 1% the sampling size would be way much higher than that. Now how does this correlate with the facial averageness study?
View attachment 526076
In order for this face to accurately represent the entire population of india (we will just assume that the population is 1 mil to meet the standards of the table) it has to be made of 16.3k layers of random faces, which is impossible on a technical level (in reality the number would be even way higher than 16.3k since it's a a population of 1.3 billion). "But what about the 2.5% and 5% margin of error?" what about them? if a study lands in any of these two areas then it's a not accurate, because lower credibility and lower sample size. So either the facial averageness study has a 1% margin of error which is impossible since you can't morph 16.3k layers of face (or more) into a single face (without it looking distorted and deformed) or it has a higher margin of error which is the most likely explanation since the only way this study is technically possible to implement is with a smaller sample size anyway, hence the high margin of error, less accuracy and credibility.
View attachment 526078
This is a picture of 225 NBA Athletes morphed into a single face, you can see how it's already starting to distort and look blurry with only 225 overlays, imagine what it would like with 16k overlays (or higher)