iblamethebrain
the ever-going process of figuring shit out
- Joined
- Dec 10, 2024
- Posts
- 677
- Reputation
- 881
(AI was used for formatting since this shitass server lacks LaTex)
Lil low effort thread, hello!
PSL space has this weird habit where people will say “looks are measurable” with full confidence, then the second you ask how exactly, the whole thing turns into mentioning incel ratios and shit arguments with someone posting a gigachad morph with zero theory behind it "do you actually find that attractive".
The irony is that the core argument is actually not that hard. People just keep mixing up three different things: subjective judgment, statistical structure, and metaphysical objectivity.
Those are not the same thing.
If enough human beings share enough perceptual biases, then facial attractiveness becomes a population-level variable that can be modeled, approximated, and eventually scored. Lets look at it.
Formulating the theory
Let x be a face.
Let ri(x) be the rating that person i gives that face (1-10 for example)
Then define the attractiveness score of that face in a target population P as:
That means A(x) is just the expected rating of face x when sampled across a defined population.
.
You can also define a harsher version:
So H(x) is the probability that a random person in the population rates the face at least 8 out of 10.
Now you suddenly have something that is crisp:
"But faces are too complex bro, there are too many variables"
Yes. Obviously. That changes nothing.
A face can be represented as a vector of features:
x=(f1,f2,f3,f4,...,fn)
where these features include things like:
The system is high-dimensional at the input level and one-dimensional at the output level. That is not a contradiction. That is normal modeling. It is how prediction works in basically every serious field.
In a very simplified form, you might write:
(meaning each trait gets a weight)
In reality it is more like:
because traits interact. Good eye area plus weak lower third does not equal the same thing as good lower third plus weak eye area. Harmony matters. Compensation matters. Synergy matters.
So yes, attractiveness can absolutely be modeled as a nearly measurable scalar even if the underlying anatomy is massively multidimensional.
Have a beautiful day!
Lil low effort thread, hello!
PSL space has this weird habit where people will say “looks are measurable” with full confidence, then the second you ask how exactly, the whole thing turns into mentioning incel ratios and shit arguments with someone posting a gigachad morph with zero theory behind it "do you actually find that attractive".
The irony is that the core argument is actually not that hard. People just keep mixing up three different things: subjective judgment, statistical structure, and metaphysical objectivity.
Those are not the same thing.
If enough human beings share enough perceptual biases, then facial attractiveness becomes a population-level variable that can be modeled, approximated, and eventually scored. Lets look at it.
Formulating the theory
Let x be a face.
Let ri(x) be the rating that person i gives that face (1-10 for example)
Then define the attractiveness score of that face in a target population P as:
That means A(x) is just the expected rating of face x when sampled across a defined population.
.
You can also define a harsher version:
So H(x) is the probability that a random person in the population rates the face at least 8 out of 10.
Now you suddenly have something that is crisp:
- higher A(x) means higher average attractiveness
- higher H(x) means more halo-level approval rate
"But faces are too complex bro, there are too many variables"
Yes. Obviously. That changes nothing.
A face can be represented as a vector of features:
x=(f1,f2,f3,f4,...,fn)
where these features include things like:
- eye spacing ratio
- canthal tilt
- jaw width
- gonial angle
The system is high-dimensional at the input level and one-dimensional at the output level. That is not a contradiction. That is normal modeling. It is how prediction works in basically every serious field.
In a very simplified form, you might write:
In reality it is more like:
because traits interact. Good eye area plus weak lower third does not equal the same thing as good lower third plus weak eye area. Harmony matters. Compensation matters. Synergy matters.
So yes, attractiveness can absolutely be modeled as a nearly measurable scalar even if the underlying anatomy is massively multidimensional.
Have a beautiful day!
