You can assign multiarm bandit trials on a lazy per user basis.
So first time user touches feature A they are assigned to some trial arm T_A and then all subsequent interactions keep them in that trial arm until the trial finishes.
careful when doing that though!
i've seen some big eyes when people assumed IDs to be uniform randomly distributed and suddenly their "test group" was 15% instead of the intended 1%.
better generate a truely random value using your languages favorite crypto functions and be able to work with it without fear of busting production
That article is mostly about speed. The following seems like the one thing that might be relevant:
> Naively, you could take the random integer and compute the remainder of the division by the size of the interval. It works because the remainder of the division by D is always smaller than D. Yet it introduces a statistical bias
That's all it says. Is the point here just that 2^31 % 17 is not zero, so 1,2,3 are potentially happening slightly more than 15,16? If so, this is not terribly important
It is not uniformly random, which is the whole point.
> That article is mostly about speed
The article is about how to actually achieve uniform random at high speed. Just doing mod is faster but does not satisfy the uniform random requirement.
additional to the other excellent comments they will become non-uniform once you start deleting records. that will break all hopes you might have had in modulo and percentages being reliable partitions because the "holes" in your ID space could be maximally bad for whatever usecase you thought up.
Not snarking, but you either have a needlessly arbitrary bar, or you've left the applicant pool up to people unqualified to gather applicants (non technical young HR, etc)
It’s LinkedIn jobs board + HN who is hiring. Everyone who applies tends to have some development experience. We interviewed maybe 30 best resumes out of the 3k applicants. Lots of people fail relatively basic python or typescript coding challenge, a few fail basic “designs an api” round. We did filter a few people in culture match but thats rare, most fail the technical rounds.
During the actual job I've never once had issues of any kind, but my brain just shuts down in interview environments. I'll forget the simplest of things that I do literally daily, and it all just ends up spiraling out of control from there. It's like there's 2 people in my brain, the regular, competent me, and then interview me who's a bumbling buffoon that I myself wouldn't hire. It's not even a pressure thing, I do fine in high stress environments, it's just specifically during interviews where things go wrong.
I hear you. I was that person once. I was able to overcome this with deliberate practice. When I was doing that, ~20 years ago, this problem was far less understood. But these days, there are many more resources to help you. Best wishes!
Yep. We pay a lot. We are a company you've heard of and we're doing well. For our systems engineering roles, we've had a hell of a time finding good people. Plenty of interviews with folks who turn out not to know basic C programming or systems level algorithms.
It seems like systems level programmers are either firmly employed somewhere else or have switch roles to an easier domain. I know I've considered going back to Python programming where I can make the same money with a lot less work.
Where "turn out not to know basic C programming or systems level algorithms" is failing some leetcode puzzle they have not touched in the last 20 years while they have been full time writing C and C++? So hell of a time not finding your definition of 'good people' would be kinda expected.
OK, yep I get that. Excuse my cynicism. True, most of us system programmers could describe, in detail, malloc and free from scratch and write a basic malloc from scratch and then know why the basic K&RT whipped up malloc would actually be quite crappy when faced with real world use.
But why assign any specific value to systemic racism vs some groups value family + education more than others. Poor Asian families suffered a lot of discrimination (and still do) but their kids do well in these tests. Ashkenazi suffered a ton of discrimination especially early/mid 20th century but still did extremely well academically. I am not even saying they are inherently smarter, I’m just saying that their value system is demonstrably different, they suffered obvious discrimination, and yet had significantly above average educational outcomes.
Why are you replying to my comment with this? It has no relevance to anything I wrote.
But since you did, I’d suggest you consider not only the value system of the victims but also that of the perpetrators and the system itself.
And also consider the history.
And consider the financial differences that often exist.
Consider the communities and their plights.
Consider destruction of cultures.
Consider the dietary and health issues that are faced.
Consider the overwhelming economic and media environments that 7 years olds grow up within and how that environment is often more impactful than parents could ever hope to be.
And, if we want to focus on biology, consider the role that vision, in particular color of skin, plays in our emotions, decision and behavior. Consider how we use color of skin to read health and emotions and intentions and how it might be harder to read those when the skin is imbued with unfamiliar tones and how on a population level, such misreads can build into mistrust and conflict.
> Why are you replying to my comment with this? It has no relevance to anything I wrote.
You emphasized systemic racism as being a major cause, but group differences can be both non-biologic and NOT related to systemic racism
> not only the value system of the victims but also that of the perpetrators and the system itself.
This assumes the answer (systemic racism) in the premise. The values of the system can be good (agency, hard work, academic pursuit, etc) and misaligned with some group. That group would then do poorly, but not because the system or its values are racist.
> And also consider the history.
I did, this is why I compared to early/mid 20th century Ashkenazi and mid/late 20th century Asians. Both were very persecuted.
> And consider the financial differences that often exist
Most asians fleeing to the US in mid 20th century were much poorer than both current as well as at that time median underperforming groups in the US.
> Consider destruction of cultures.
If anything, current underperforming groups (eg african americans) are famous for having a lot of cultural products. This is where they thrive.
> Consider the dietary and health issues that are faced.
Again, both ashkinazi and asian groups suffered famines + serious malnutrition. Very few in american disadvantaged groups are in danger of starvation or serious malnutrition.
> Consider the overwhelming economic and media environments that 7 years olds grow up within and how that environment is often more impactful than parents could ever hope to be.
Everyone has access to all the same media. There is a significant effort (which I agree with) to over-represent underprivileged groups as successful heroes in modern TV/etc. Parents have significant influence on which media mix is consumed and what counts as "success." Both asians and ashkinazi were represented very negatively in the media mix of mid 20th century, yet they thrived. Nigerian american diaspora today thrives as well (unlike most other african american groups).
> And, if we want to focus on biology, consider the role that vision, in particular color of skin, plays in our emotions, decision and behavior. Consider how we use color of skin to read health and emotions and intentions and how it might be harder to read those when the skin is imbued with unfamiliar tones and how on a population level, such misreads can build into mistrust and conflict.
I specifically didn't focus on biology, but Ashkenazi were clearly targeted based on how they looked. Caricatures of "the Jew" were popular and everywhere in early to mid 20th century Europe. People perceived them especially as untrustworthy. Asians are also obviously and easily identified by a quick look at their face. South asians also have "brown" skin color, that is very similar to that of disadvantaged groups in the US, yet they do well academically/financially/etc. Most people can't tell apart nigerian americans from other african americans, yet nigerian americans tend to do well.
---
In all of this i'm not saying hardship doesn't exist, or that racism doesn't exist, or that differences are biological. I am saying that there is a confounding factor that is essentially bigger then all of this. I think this confounder is "culture/value system" of the group. Not all cultures/value systems are equal, not all of them lead to the same outcomes, these differences are not racist.
> You emphasized systemic racism as being a major cause, but group differences can be both non-biologic and NOT related to systemic racism
Firstly, I did not. I simply pointed out that racism can be systemic without individual contribution.
Secondly, the fact that racism can arise from a third option (neither non systemic and non biologic) does not change the fact that the article did not accuse teachers of racism.
And that’s where my initial comment stopped.
As for the rest, I had hoped it would have made it clear that the issue is too complex to unravel and try to assign blame or cause. The factors are too nuanced, the history too complex, the societies and neighborhoods too diverse.
But for some reason, you seem to have a need to find a cultural- or values- based factor for differing outcomes. You can do that if you wish. I won’t partake though. (Still not sure why you even chose my comment to initiate such an attempt)
Edit: I should state that the experience of non-immigrant blacks, in the US, is entirely incomparable to that of either Asians or Ashkenazi or even Nigerians.
And looking at actual outcomes in the US it’s easy to see that the truth is different. It’s not even white kids that come up on top, it’s mostly Asian kids (and before that Ashkinazi kids). It’s not because they have some institutional privilege. It’s because culture matters and valuing smarts and education is important not just for test taking but also for benefiting the society long term.
If you do merit based acceptance into programs then obviously it will have a different demographic makeup than population at large. We can discuss the causes of this elsewhere, but obviously test/school performance varies significantly by ethnicity today in the US.
We want to help every person to feel in control of their healthcare: to know what's going on with their body and to not feel alone.
MDandMe (https://mdme.ai) builds generative AI technology that allows people to make sense of their health at all steps along their healthcare journey: from a conversational symptom checker when you first feel unwell, to a community where you are matched to others with near-identical medical histories, to follow-up conversations where you can have diagnostic reports, lab values, and doctors notes explained in plain language.
We are looking for a Mobile SWE (React Native, iOS + Android): over 93% of our traffic comes from mobile devices. Right now we are a React/NextJS webapp and are currently building a React Native mobile app. https://swift-step-f49.notion.site/Mobile-SWE-06f94fd945bd4f...
We want to help every person to feel in control of their healthcare: to know what's going on with their body and to not feel alone.
MDandMe (https://mdme.ai) builds generative AI technology that allows people to make sense of their health at all steps along their healthcare journey: from a conversational symptom checker when you first feel unwell, to a community where you are matched to others with near-identical medical histories, to follow-up conversations where you can have diagnostic reports, lab values, and doctors notes explained in plain language.
We are looking for a Mobile SWE (React Native, iOS + Android): over 93% of our traffic comes from mobile devices. Right now we are a React/NextJS webapp and are currently building a React Native mobile app. https://swift-step-f49.notion.site/Mobile-SWE-06f94fd945bd4f...
