wais5psl
Iron
- Joined
- Mar 10, 2026
- Posts
- 52
- Reputation
- 52
For context:
Looksiq lets people vote who is more attractive and is based on IRT (Item-response-theory).
Since release we have accumulated more than 250k votes and nearly 1000 people have participated by uploading a selfie! I want to express my gratitude as I didn’t expect the community to be so involved and supportive.
**But now lets get to the interesting part!**
1. The inter-rater-agreement sits at approximately 90%. This means that 90 out of 100 Times people agree regarding which individual is more attractive. The number may even be 'deflationary' considering that there are very close matches where different opinions are to be expected.
2. People who upload multiple pictures received extremely similar scores:
This is exciting as it demonstrates that there is a latent variable. So even if the conditions were different (lets say different lighting) the score stayed consistent.
3. Looksmaxxers and normies vote similarly:
I invited many normies to try the website and to my surprise their voting patterns didn’t diverge strongly. Many people adhere to the proposition that Looksmaxxers have different ideals which doesn’t seem to be true.
conclusion: The data suggests that attractiveness is a latent variable and that people regardless if they are looksmaxxers or not have the same beauty standards!
Future investigations: I will add an advice feature that allows people to give feedback. I will standardize it utilizing machine learning. The question is if advice given by lots of people actually is good advice.
Looksiq lets people vote who is more attractive and is based on IRT (Item-response-theory).
Since release we have accumulated more than 250k votes and nearly 1000 people have participated by uploading a selfie! I want to express my gratitude as I didn’t expect the community to be so involved and supportive.
**But now lets get to the interesting part!**
1. The inter-rater-agreement sits at approximately 90%. This means that 90 out of 100 Times people agree regarding which individual is more attractive. The number may even be 'deflationary' considering that there are very close matches where different opinions are to be expected.
2. People who upload multiple pictures received extremely similar scores:
This is exciting as it demonstrates that there is a latent variable. So even if the conditions were different (lets say different lighting) the score stayed consistent.
3. Looksmaxxers and normies vote similarly:
I invited many normies to try the website and to my surprise their voting patterns didn’t diverge strongly. Many people adhere to the proposition that Looksmaxxers have different ideals which doesn’t seem to be true.
conclusion: The data suggests that attractiveness is a latent variable and that people regardless if they are looksmaxxers or not have the same beauty standards!
Future investigations: I will add an advice feature that allows people to give feedback. I will standardize it utilizing machine learning. The question is if advice given by lots of people actually is good advice.
