What the resume-in-pizza-boxes channel offers over, say, the resumes-in-email channel, is a lot fewer resumes. You've got some intention signalling in the RIPB version (it's expensive in terms of time and other committments), but low informational signalling otherwise as to overt qualifications, outside some general concepts (hustle, motivation, time availability / opportunity costs, charisma, confidence, or simply having heard of the trick somewhere).
There are a bunch of selection hueristics whose primary benefit is they eliminate a large component of the choice process. You'd do fairly well in simply randomly sampling from amongst the submissions, or applying some very rudimentary filters to eliminate the most obviously bad choices.
This is why mechanisms such as racial, gender, age, and cultural bias are so common. They're easy to apply, aren't entirely wrong (relatively homogeneous groups tend to operate better than nonhomogeneous ones, though there are many arguments about resilience and creativity/diversity), and hugely reduce the decisionmaking costs of the selector.
Unfortunately, knowledge / infotech / creative work hinges hugely on intangibles which are quite difficult to effectively communicate, especially to someone (say, an HR flunky) not intimately familiar with the field.
Toss in a ton of employer biases and control incentives (e.g., anti-poaching, anti-union, and similar concerns) and the situation starts getting messy quickly. Sure, there might be some alterative opportunities created as a result, but these tend to have far less commercial/financial potential and viability.
I'll accept cultural homogeneity having worked in different cultures myself. But if you're seriously suggesting that racial, gender and age groupings are correlated with performance and subconsciously okay, you're going to have to provide a very convincing argument.
On an intellectual basis, I can acknowledge, though, that high-affinity classifiers -- ones with a strong, clear, and generally readily determined distinguishing capacity -- are frequently used. Language, speech, religion, gender, dress, custom, and more. People are inherently clannish. You'll see this not only in tech hiring, but in ethnic conflicts: Iraq, schoolyards around the world, Nanking & Korea in WWII, Syria, prison populations, the Stanford Prison Experiment, Tunesia, the various ape species actors in the Planet of the Apes cafeteria, Columbia, Yemen, England vs. Ireland, England vs. Scotland, Flemish vs. Walloon, Basque vs. Spanish, various subreddits, Rwanda, Yugoslavia, Israel-Palestine, and many, many more.
A principle problem of any seggregation regime is that it tends to require oppressive, often violent, occasionally genocidal enforcement to be sustained. Which is a key reason to seek integrated rather than segregated societies.
However that integration is going to have to be sustained against an apparently very-deeply ingrained resistance to doing so. And simply demanding conformance is likely to be at best only partially successful. The record of the US in that effort, well over 50 years after the Civil Rights movement, and over 150 years after the Civil War and just shy 150 years after the 14th amendment, illustrates this.
I'm not claiming homogeneous societies are better or preferable. I'm strongly suggesting they're easier to establish and sustain, and that they tend to self-organise as such.
I mean you may claim no one can force this heterogenous utopia but it is most certainly the future.
Or as decisionmaking: https://hn.algolia.com/?query=dating%20choice&sort=byPopular...
I'm reminded of the hiring manager that randomly tosses into the shredder 90% of incoming resumes. "I don't want to hire someone who's unlucky!"
I'd suggest a below-bar elimination check (e.g., spelling, gross formatting), and an high-level check in case, say, Ted Ts'o happens to be applying. Take the bulk of largely undifferentiated but not obviously bad resumes, shuffle them, and select randomly from them. Pick your sample, apply your assessments, and proceed.
The low-cost primary selection, and the unbiased secondary selection, leave the expensive selection process (resume screen, phone screen, interviews, background / portfolio assessment, etc.) with a tractable number of candidates to deal with.
Aeon's "How to Choose" is an excellent article on the use of randomness: https://aeon.co/essays/if-you-can-t-choose-wisely-choose-ran...
Someone willing to try this is clearly highly motivated in their job search. Which is one of several axis of qualifications you'd wan to consider. Depending on the position sought (I've not RTFAd), the gumption could be a positive indicator for, say, marketing or leadership (charisma, initiative, risk-tolerance, confidence) though quite likely flat or negative for programming / knowledge-work (humility can matter a lot there).
The economics of information, signalling, signalling costs, proxies for information, quality assessment, etc., are a fascinating and IMO under-studied area. In another tab I'm going through the American Economic Association's annual Papers & Presentations issues looking for topics covered. There are a few on information (early 1970s) that I've found yet. But overall, horribly broken.
It's a bit like the Xerox-generation effect -- if you take an image and photocopy it repeatedly, you'll get an increasingly high-contrast image (a similar effect comes from playing with contrast settings in a photo editor such as the GIMP). But what's lost is the subtlty and nuance. For images which present well in high-contrast (say, the Eiffel Tower against the sky), this isn't terrible. Monet's Rouen Cathedral, Facade (Morning effect) not so much.
https://commons.m.wikimedia.org/wiki/File:Claude_Monet_-_Rou...
The children's game of "telephone" (in which a message is passed by voice from person to person around a circle) is similar.
I'm starting to wonder what examples there are of error-correcting high-quality, high-relevance information transmission systems within biology and elsewhere.