My experiment: guessing ethnicity from photos

Domen

Misico dux Vandalorum
Joined
Mar 27, 2009
Messages
8,088
Location
Doggerland
So I did an experiment:

I gathered 912 photos of 912 Germans and Poles, and I asked people to label each photo with "G" or "P", to see if they can distinguish ethnicity.

My sample of 912 was probably as representative as it could be (I can't imagine a more representative sample, but I won't reveal why yet).

So far only one American guy - who is of Basque ancestry, who is a teacher, and who travels a lot (he claims that he visited Austria, Bulgaria, Czech Republic, Germany, Hungary, Lithuania, Netherlands, Poland, Romania, Slovakia, Slovenia and Spain) - volunteered to solve my poll.

He omitted several individuals, but generally he labelled most of them (around 900).

His results were as follows:

Out of all Germans, he identified 56.608 percent as Germans and 42.952 percent as Poles.

Out of all Poles, he identified 54.625 percent as Germans (!) and 44.323 percent as Poles.

I'm not sure what to think of these results. :lol:

On one hand he identified a lot of Germans wrongly as Poles, but on the other hand he identified the majority of Poles - wrongly - as Germans.

=============================

Does anyone want to try if you can do better ??? :D

Here are the photos (divided into two files, each set includes members of both ethnic groups):

http://postimg.org/image/dc5r8yp4d/full/

http://postimg.org/image/6ajrmpreb/full/
 
There are two kinds of Poles: Russians pretending to be German and Germans pretending to be Russian. No surprises there.
 
I'm not sure what to think of these results. :lol:


To rephrase your question, you want people to do your dirty work of statistical data analysis. ;)
One interesting result would be whether your test person is better than a random number generator with the same bias towards Germans. You need to calculate the binomial distribution for the probability to get n people right. Then you can tell how many standard deviations he is off the expectation value. If it's something like >3, he may claim that he can sometimes tell Germans and Poles apart. But from your description it seems like it's <1.
 
Yes its really easy to tell white people apart.

No, not really.
 
Gigaz said:
One interesting result would be whether your test person is better than a random number generator with the same bias towards Germans. You need to calculate the binomial distribution for the probability to get n people right. Then you can tell how many standard deviations he is off the expectation value. If it's something like >3, he may claim that he can sometimes tell Germans and Poles apart. But from your description it seems like it's <1.

Thanks! Do you have the right formula for this calculation at hand?

bhavyy said:
Yes its really easy to tell white people apart.

No, not really.

I told you in another thread, "white people" live mostly in the USA.

Europe has never seen them. :)
 
Thanks! Do you have the right formula for this calculation at hand?

Making some assumptions about the type of distribution in your data, but a student's t-test (see the one sample location tests section) is probably appropriate and will get you started.

Given the near 50/50 rate of misidentification, I'm betting you are going to see he couldn't tell the difference. Unless some of the pictures showed them with national flags.
 
Comparing faces of two neighbouring ethnic groups is generally a poor idea, unless it is a matter of melanin. A lot of mixture between Germans and neighbouring Slavic peoples has occurred, so physical similarities are bound to occur. Hairstyles and clothing, cultural factors, probably give more away.
 
This title is approximately 100% better if you read is at "guessing ethnicity from potatoes".
 
I know instance of Turks who misidentified Bulgarians as fellow countrymen!

The 1980 coup in Turkey was probably committed by Bulgarian imposters.
 
obama%20sunglasses%20helicopter.jpg


Spoiler :
Irish


080519_mccain_adler.jpg


Spoiler :
Panamanian
 
Torvegieter said:
A lot of mixture between Germans and neighbouring Slavic peoples has occurred, so physical similarities are bound to occur.

Mixing of course, has been taking place, but also already ancestors of both groups before mixing could be similar.

Torvegieter said:
Hairstyles and clothing, cultural factors, probably give more away.

Traditional old Polish hairstyle, chupryna: :D

086de2d22802d1e50b175dba1d5c0dcd.jpg


Or maybe it is Ukrainian-Russian originally.
 
You collected over 900 photos of Poles and Germans? Wouldn't that take a huge amount of time? And you managed to get someone to actually go through hundreds of photos and guess which is which? What's the point?
 
NovaKart,

You collected over 900 photos of Poles and Germans? Wouldn't that take a huge amount of time?

No because I found an easy way to collect them, it took me several hours, IIRC.

BTW, I did that last year, about a year ago, but just recently someone has finally decided to solve the poll.

And you managed to get someone to actually go through hundreds of photos and guess which is which?

Yes, after one year one guy decided to do this.

What's the point?

I'm not even sure, yet.

And what's the point of us all posting in "Off Topic", which is about nothing-everything, as its name says. :)

==============

JollyRoger - good job! Continue posting comparisons. :D
 
Back
Top Bottom