Despite what you may have heard, the Okcupid blog post, "Your Looks and Your Inbox," does show a substantial messaging premium for attractive males (Rudder, 2009)

Since bluepill advocates seem to be fixated on Okcupid's blog post on attractiveness and messaging rates, a more disciplined look at the content is well overdue.

First, I'll start with what is arguably the most abused portions of the blog post: the messaging and attractiveness density histograms:

for male messaging: https://cdn-images-1.medium.com/max/800/0*aiEOj6bJOf5mZX_z.png

and for female messaging: https://cdn-images-1.medium.com/max/800/0*aWz0dYzuUR7PO3dP.png

These images are often disembodied from the rest of the blog content and spewed across reddit as "atomic bluepill" failevidence to counter redpill and blackpill claims.

The problem is the blog post clearly states this about the density histograms:

The information I’ll present in this post is not normalized

This is crucial to interpreting the histograms. It's clear the messaging plots are simply showing the total number of messages received by each looks rating as a proportion (%) of the total number of messages sent out on their platform, but because no normalization was performed, the messaging data is raw and uncorrected for the number of individuals at each rating level. 100 messages going to 100 different individuals is much different than 100 messages going to 10, but you can't even infer that level of granularity with the data (no absolute numbers provided).

Thankfully, the blog author did include a corrected graph, and here it is:

https://cdn-images-1.medium.com/max/800/0*rRhMB4YoU-HURGeE.png

Sure, the female recipient graph is exponential while the male recipient graph looks cubic, but note the scale is in multipliers and, unfortunately, absolute numbers were not given anywhere in the blog post. It is almost certain (based on, for instance, Hitsch 2006 and 2010) that there is at least an order of magnitude more messages being received by female recipients than male recipients, such that the gender-controlled multipliers conceal the likely massive disparity.

It should also be pointed out that the messaging best fit trend line for male recipients is similar to what Hitsch 2006 described before binning out men in the top 5% of looks. Hence, it is entirely possible the data, by nature of the 0-5 rating system, obscures a winner-takes-all "superstar effect" Hitsch and colleagues identified in their dataset.

The Okcupid blog concludes by showcasing the reply rates data, which is consistent with expected trends.

tl;dr: Overall, the entire blog post is consistent with the well-supported observation that attractiveness is the most robust predictor of initial romantic interest.

[–]neomancrAt-risk Sperg0 points1 point2 points 8 years ago* (7 children) | Copy Link

This is just a "nothing to see here" post. You still haven't established that women all have the same tastes. Why would you even imagine that would be the case when we can't even establish what their favorite food is.

You can determine that women may like sweeter water, but you can't claim that women must then only want the sweetest foods.

Yes, we have found that women respond similarly to drugs and even to fire, but that is no reason for us to presume that they respond the same way to male aesthetics. Attraction is much less direct.

All the tests that I've seen trying to conclude homogenous female attraction make the same mistake and are trying to prove something that is already presupposed from the start.

It is always amorphous androgenous blob versus human with whatever trait we want to prove is superior. All were proving is that women do not like amorphous androgenous blobs over and over again.

As far as the graph you can clearly see that the whatever you can glean about females would have to be much more so for males.

Here is an article that demonstrates how much the media impacts our tastes:

https://www.livescience.com/46929-internet-access-attractive-features.html

Perhaps if we exposed all women to he same media we can force them to prefer the same Chad but that is pure conjecture

(this is my best attempt at capturing the tone of this sub)

Added:

Here is a more straight forward chart that would refute the claim that there is some hidden aspect suppressed by the multiplier.

https://cdn-images-1.medium.com/max/1690/0*cWWqWB6MMEtu6SvX.png

Although we have to take into consideration that fact that more generally attractive people will be generally better at initiating conversations, nonetheless the chart is still flatter than for men. You can also see more distortion along the curve which indicates less homogenous preferences.

[–]SubsaharanAmericanshitty h-index[S] 4 points5 points6 points 8 years ago* (6 children) | Copy Link

You very clearly do not understand the arguments I am making. I never said women have homogenous tastes. In fact, I very specifically indicated that they don't (hence my explicit denial of lookism determinism), but that nevertheless, a model that's insensitive to picking up subgroup heterogeneity because everything is averaged (e.g., a simple linear model) can still capture overall trends. Do you even know what a regression analysis/model is? I suggest you read up on it before replying further. Learn the differences between linear, logit, probit, fixed effects and mixed effects, for starters. These are very basic statistical analyses performed in just about all of disciplines of science.

The reply rates no way gives any indication of the absolute number of messages sent, or at least that's what I think you're referring to when you claim I claim "hidden aspect suppressed by the multiplier"

The argument you're trying to make is that some subgroup or diffuse heterogeneity precludes any statistical analyses. Except for the fact that if this were true then:

there would be poor correlation of ratings between different independent observers used in the studies for a single final rating (usually a central tendency metric such as mean) to be useful (this is measured by the alpha index, but the way), and
the independent observer consensus rating would have no predictive value, and
the correlation between consensus ratings and outcomes wouldn't be reproducible

Except the studies show none of these are true. That is, the ratings correlate, they are predictive, and all studies employing such a methodology come to the same findings.

Facial masculinity correlates (or "media halo effects" on the attractiveness of facial masculinity) are irrelevant to the big scheme here. For instance, if you have 100 independent raters, and 80 like highly masculine faces from the internet exposure (or whatever), 20 like feminine faces, then sure, the 80 will drive the final "consensus" score and a high alpha. But if such a metric was as volatile as you seem to suggest, you wouldn't predict that it could generalize in such a way as to be predictive in online dating and in speed dating. If you were to argue the speed dating and online dating cohorts also contained similar rates of "media halo effect" exposed targets, then for this to harm the external validity then you must also argue that real world targets systematically deviate from the preferences of both the consensus raters and the targets, but you have not shown such. Neither does your pop psy shitty article summarizing the findings of a survey performed in El Salvador (yeh, real generalizable) and published in a low level open access journal

[–]neomancrAt-risk Sperg0 points1 point2 points 8 years ago* (5 children) | Copy Link

You very clearly do not understand the arguments I am making. I never said women have homogenous tastes. In fact, I very specifically indicated that they don't (hence my explicit denial of lookism determinism), but that nevertheless, a model that's insensitive to picking up subgroup heterogeneity because everything is averaged (e.g., a simple linear model) can still capture overall trends. Do you even know what a regression analysis/model is? I suggest you read up on it before replying further. Learn the differences between linear, logit, probit, fixed effects and mixed effects, for starters. These are very basic statistical analyses performed in just about all of disciplines of science.

The reply rates no way gives any indication of the absolute number of messages sent, or at least that's what I think you're referring to when you claim I claim "hidden aspect suppressed by the multiplier"

The argument you're trying to make is that some subgroup or diffuse heterogeneity precludes any statistical analyses. Except for the fact that if this were true then:

there would be poor correlation of ratings between different independent observers used in the studies for a single final rating (usually a central tendency metric such as mean) to be useful (this is measured by the alpha index, but the way), and

the independent observer consensus rating would have no predictive value, and

Nope, it would only show less predictive value versus men which it does.

the correlation between consensus ratings and outcomes wouldn't be reproducible

No there can be trends even with more variegated base data. We've been trying to shape what women are attracted to for a hundred years. So of course that will have some influence. If it didn't why would we go through the trouble? It was so important that we had to pass several laws and product codes to control media representation of ourselves from the start.

Beyond that

Refer to this:

https://www.reddit.com/r/Braincels/comments/81kltc/ok_cupid_studies_proving_how_women_have_much_more

Look at the messaging curve. Then refer back to the above linked analysis.

The reason why r is "not shown" is already explained. The studies are looking for the results they're trying to prove. If you reduce the factors to only looks you will get a general tendency but you would have to compare the weight of that value to other traits as well.

[–]SubsaharanAmericanshitty h-index[S] 5 points6 points7 points 8 years ago* (4 children) | Copy Link

[–]neomancrAt-risk Sperg0 points1 point2 points 8 years ago (3 children) | Copy Link

[–]SubsaharanAmericanshitty h-index[S] 1 point2 points3 points 8 years ago (2 children) | Copy Link

[–]neomancrAt-risk Sperg1 point2 points3 points 8 years ago* (1 child) | Copy Link

[–]SubsaharanAmericanshitty h-index[S] 3 points4 points5 points 8 years ago (0 children) | Copy Link

Banned this commenter for 30 days for Rule 2-3 violations after finding out he's a known troll who spergs at length about random gibberish. See https://www.reddit.com/r/BlackPillScience/comments/8blrj8/ok_cupid_studies_proving_how_women_have_much_more/dxd6dex/ for the abjectly frustrating dialogue that took place prior to this discussion for full context.

The slopes (or, rather, the presumed Beta coefficients assessed via eyeballing) of the gender-stratified univariate analyses in OKcupid's analyses indicate both genders are responsive to increasing attractiveness but that men are more so. Perhaps. This should be viewed in full context, however. Namely:

that men are also more generous when it comes to ratings (this is a reproducible finding, Eastwick 2018 actually has a recent pub also showing this)
OKcupid did not bin the rating data appropriately to be sensitive enough to pick up on any uniquely strong premium in the very top percentile of men (Hitsch described that in their paper, before binning out the top 5%, the trajectory was smooth and linear); it looks like OKcupid binned the data into categories of sixths, then connected the dots -- not sufficient
Descriptive stats on the data, including absolute numbers and number of individuals who didn't receive ANY messages, are missing from OKcupid's blog post. In Hitsch data: ~56% of the men did not receive a first-contact message, ~20% of women didn't, there was a vast disparity in the central tendency measurements as well (median 0 mean 2 messages received for men vs median 4 mean 11 for women). Hence, why I say OKcupid's line convergence is misleading without knowing more details about their data.

It should also be noted, OKcupid's modeling makes the same assumptions as the peer-reviewed attraction literature in the sidebar, then it makes other assumptions that are likely wrong. The commenter was sperging at the fact linear models are insensitive to subgroup heterogeneity, without realizing this was OKcupid's approach to their data with a univariate model stratified only by gender. All interpretations in Okcupid's blog was derived from this univariate model uncontrolled for other covariates. No multivariate analyses were performed to see what effect other variables, such as race/ethnicity, might have on the coefficients (in fact, Okcupid split out the race/ethnicity stuff into a separate blog posts again using a univariate approach).

Comments