Correct, and algorithm shouldn't use a protected class as an input. But at the same time, simply pointing out that a inequitable outcome is produced is not evidence of bias.
Also, dual citizenship is not a protected class last time I checked.
> When AI criminal risk prediction software used by judges in deciding the severity of punishment for those convicted predicts a higher chance of future offence for a young, Black first time offender than for an older white repeat felon.
Younger people are more likely to re-offend than older people. Remove race from the situation entirely, and this is still the expected result. There's zero reason to think this algorithm was based with respect to race.
If you read deeper than a one line description you'll see:
1. Even after accounting for criminal history, recidivism, age and gender, black defendants were still scored as much more likely to re-offend.
2. It incorrectly predicted that black defendants would re-offend much more frequently than white defendants.
3. It incorrectly predicted that white defendants would not re-offend much more frequently than black defendants.
Considering those three things and that they have refused to give any details about how the algorithm works, I see zero reason to give them the benefit of the doubt. Fire them until they can demonstrate that the algorithm is not in fact biased like it appears to be.
The score also takes into account answers to questions like whether the defendant's parents were separated or whether their parents were ever arrested. Those things are completely out of the defendant's control and are highly correlated with race. Even including those things in their score is damning.
Phrasing it in the way you do is misleading. The defendants labeled as high risk and low risk were just as likely to re-offend. To put this in simpler numbers
* Out of 100 white people, 5 were labeled high risk.
* Out of 100 black people, 20 were labeled high risk.
* Out of the 5 white people labeled high risk, 4 re-offended.
* Out of 20 black people labeled high risk, 16 re-offended.
In either case, someone labeled high risk had the same likelihood to re-offend: 80%.
"It incorrectly predicted that black defendants would re-offend more frequently than white defendants." This is technically correct, but not because the algorithm was bad a predicting rates if re-offending. It's because there was higher rates of re-offending. The likelihood of re-offending among someone labeled high risk is the same.
This kind of objection seems like a blanket rejection of any system that produces an inequitable outcome. But the reality is that rates of re-offending is not equal. Even a perfectly accurate prediction of re-offense is going to predict higher rates of re-offending among men. Because men re-offend at higher rates. This isn't sexism.
That doesn't mean we shouldn't recognize the disparate impact of incarceration on underprivileged people. But simply concluding bias due to inequitable outcomes is simplistic.
Your analysis above is a prime example of how dangerous statistics can be when not properly considered. Even if we took your numbers above as truth, it actually does not indicate that the system isn't biased. In order to determine that you'd need to know what percentage of each group _not_ labeled as high risk never re-offended. You're only analyzing the true positive rate without taking into account the false negative rate. You'd need to know how many people in each group that weren't labeled went on to re-offend.
Thankfully, the article I linked to looked at both:
White African American
Labeled Higher Risk, But Didn’t Re-Offend 23.5% 44.9%
Labeled Lower Risk, Yet Did Re-Offend 47.7% 28.0%
When you are doing crystal ball voodoo based on stuff that has no connection to the defendant's choices and no direct connection to criminality like whether or not their parents were separated and whether their parents were ever arrested, then you're damn right there should be a blanket rejection of any inequitable outcome.
I'm not even convinced that you should be allowed to make a decision based on stuff like that even if it is somehow equitable.
> no direct connection to criminality like whether or not their parents were separated and whether their parents were ever arrested
Again, not right at all. These things heavily correlate with crime. A broken home is the primary indicator of someone's future success, even over a 'better' but broken home.
I'm male, and not a rapist, but having a penis heavily correlates with rape. I suggest you do not pick me, or any male, to watch your children. Sure it's rude to the innocent, but oh well, a little rudeness vs potential harm.
> I'm not even convinced that you should be allowed to make a decision based on stuff like that even if it is somehow equitable.
Then nobody will ever follow the law. If it's illegal for me to reject a potentially bad babysitter for something I know about them that could risk my child's safety I'll happily lie and say I didn't like their haircut.
You'd get further if you tried to identify these people and treat them better - grants to move out of bad neighborhoods, to get educated, to get pardons for unrelated crimes, to get counseling for abuse, etc - than with this "wrong unless it's 100% identical, equity-of-outcome" thing.
We're talking about an algorithm being used as part of the court system to determine whether or not someone rots in jail, not about how you choose your babysitter.
It needs to be held to a higher standard and punish people based on their actions, not based on what their parents did.
No, you're talking about making it illegal to use the things you know about someone or something to make a fully qualified decision. About a babysitter, a potential criminal, an immigrant, whatever.
> It needs to be held to a higher standard and punish people based on their actions, not based on what their parents did.
And that's not what actually happened. A guy got arrested on bad data and then released, we know some kids might be at risk because they're from broken homes, etc. Nobody was jailed for their parent's actions, and nobody even proposed it. Some trends are worrying, but you act like they're the intent of the entire system not just some scammy products that a company is pushing.
Voting machines, for instance, should all be burned. But I understand that most people don't know why and support them for convenience. I think they're anti-democratic but I don't think people are evil for using them. You should try to get a similar perspective.
I am not talking about that. You are misinterpreting my statements. "stuff that has no connection to the defendant's choices". Defendants only exist within the court system.
> Nobody was jailed for their parent's actions
Not jailed, but they are being kept in jail because of them. That is the same thing.
I am not claiming that anyone is evil. Just that this is unjust and that they need to stop doing this. You are putting words in my mouth.
> Fire them until they can demonstrate that the algorithm is not in fact biased like it appears to be.
What do the non-black box findings from the area show? If blacks are a poor demographic in the area it might be true. (Crime tracks poverty, not race.)
But yeah, certainly don't pay anyone for, or use, a black-box algorithm.
> The score also takes into account answers to questions like whether the defendant's parents were separated or whether their parents were ever arrested. Those things are completely out of the defendant's control and are highly correlated with race. Even including those things in their score is damning.
Nope. That's perfectly fair to look at. For instance, a broken family is another predictor. It's not fair, but being a victim increases your chance to offend. Similarly, the number one predictor of child sex crimes is to have suffered them yourself. If you have a child and are picking a babysitter, skip the one who was molested.
fwiw, those things don't correlate with race, they correlate with poverty which correlates with race. Broken families are more likely to be poor and abusive.
> Therefore the only cost estimate we can rely on is what we have actually managed thus far.
Correct. And we've managed to bring the cost of nuclear down to 1-2 billion dollars per GW when building plants at scale [1]. This was done during the late 60s and early 70s when we built lots of nuclear plants at once [2].
> These results show that there is no single or intrinsic learning rate that we should expect for nuclear power technology, nor an expected cost trend.
Right, and currently we are barely building any nuclear reactors, so the prices people are quoting as the $/MWh are the prices when you have zero economies of scale.
Meanwhile the prices people are quoting for solar are with massive economies of scale at this point.
We need economies of scale, and I'm not talking about 1970s scale, I'm talking about scale.
I'm not sure what you're talking about. Nuclear gets cheaper when it's built at scale. The price history for nuclear demonstrates this: it was considerably cheaper when many plants were built in parallel. It's expensive now, because it's being built at very small scales.
1970s scales work fine. It was built at a cost of 1-2 billion dollars per GW, which is very competitive versus solar and wind + storage.
Yes, that is my point. Not sure why you disagreed with me in your first comment that currently quoted prices for nuclear are with effectively no economies of scale.
Maybe you misunderstood and thought I was saying that you can't have economies of scale with nuclear? If so, that is pretty much the opposite of my intended meaning.
What do you mean? Over the course of a decade and a half we brought the share of nuclear generation to 20%. That's way more impressive than what intermittent sources have achieved.
Persistently near-exponential growth over multiple decades isn’t impressive to you? If this carries on, peak PV output alone will exceed 100% of current total power demand by 2030.
(Naturally this means I think the growth seen since 1992 won’t last until 2030, but the point remains)
Assuming that exponential growth will be steady is a very risky assumption. Where's my single-core 100 GHz CPU? Moore's law says we should have that by now.
I'm not sure I follow. The fact that a nuclear grid doesn't require storage is precisely why it's so much better. In order to make renewables viable we'd need to redirect a huge portion of our storage capacity away from electric vehicles and towards grid storage instead.
> That is, if you assume we'll need nuclear power, you're basically assuming we're not going to solve the climate crisis.
How on earth does this make sense? The country that has come closest to solving the climate crisis (France) is heavily nuclear. Their carbon intensity of electricity is 1/7th that of Germany.
It's always funny to me how proponents of nuclear power use France as their example while ignoring that France is trying to dramatically reduce the amount of nuclear power they have because it's too expensive. If you're going to use them as an example, let's use them as a full example instead of just trying to paint a rosy picture on a nuanced subject.
There's just so much misinformation about nuclear out there it's kind of crazy.
Another claim I see regularly: Angela Merkel decided on the German nuclear exit as a response to Fukushima and she permanently shut down all reactors.
When in reality Merkel tried to draw out and prolong a nuclear exit that was already decided and ratified in the early 2000s and Germany still has fission reactors running to this day.
The GP point is valid. France production is much cleaner thanks to nuclear energy. Whether nuclear energy being "too expensive" is actually a relevant problem in this discussion depends on your ideas about the role of government, importance of the CO2 reduction, importance of big private profits and so on.
The paradox of tolerance is simply factually incorrect when it comes to speech. The existence of intolerant speech doesn't exclude anyone. People who are offended don't have to listen. The paradox of tolerance only exists when people or groups are taking action to exclude others. And we don't have blanket tolerance for actions, only for speech. Someone is free to talk the talk of intolerance, but the moment talk turns into action the law swoops in and takes them away.
In short, the paradox of tolerance doesn't hold true at all when we're talking about blanket freedom of speech. If we were talking about anarchy, blanket freedom not just of speech but of actions, then yes the paradox of tolerance comes into play. Some
This kind of reaction is what phrases like "it's okay to be white" and "all lives matter" are meant to elicit. It's exactly what trolls want you to do. The other 90% of the population is appalled. Is it not okay to be white? Do some lives not matter?
There are people explicitly saying that no, it's not OK to be white - that if you're white you are automatically racist, and automatically guilty for what happened in the past. White Fragility is an example.
"It's OK to be white" isn't just a dogwhistle. (It may be that, but it isn't just that.) It's an explicit rejection of that kind of "all whites are guilty racists" baloney.
That's what makes it such an effective troll. If it was simply obviously bad from all conceivable contexts it wouldn't work. But, you are still being trolled. And, you are playing right into their game. They are getting you to sing their slogan and laughing at you the whole time.
Yes, there are some people out there in social media who have taken to hating white people for being white. That's not OK. Call it out. But, don't dance for the trolls in the process.
There's the group of anti-white haters who need to be called out. There's the vast majority of people who don't want any form of racial hatred. And, then there's the white supremacist trolls (and just plain trolls) who want to trick the vast majority of people into fighting themselves and in the process make it look like anti-white hatred is much, much more common than it really is.
"It's OK to be white" is a troll campaign that is actively being used to bait people into fighting each other and to have both sides come out thinking the other is far more racist than they really are.
> I'm not sure where the logic is here, you are making an assertion without any supporting evidence.
You might want to take your own advice. You did the exact same ting, make an assertion without supporting evidence, in your previous comment.
> What sort of logic can you provide to back up your assertion? Any such reasoning must also include actual numbers and costs
Nuclear power was built with an average cost of $2-3 billion dollars per GWe of capacity when nuclear was built at scale during the late 1960s through much of the 1980s [1]. The US was building nuclear plants at a rate of several plants per year, rather than several plants per decade. The same pattern holds true in France. During the 1970s, 80s, and early 90s, plants of the same design were built in serial production. Now, when they're building plants one or two at a time it's more expensive.
There's a clear pattern of cheaper plants when built at scale, and this pattern holds true across the two main nuclear power producers.
I wasn't the one who said my assertion was true because of "logic"
You seem to be backing up my point with your claims. Nuclear has consistently gotten more expensive over time. The designs you point to from the 1970s will never be built again, because nuclear engineers realized their flaws and don't want to build them again.
With each new generation of nuclear, it gets more expensive, not less. As we refine the designs, we spend more per GW, not less. And not just a little bit more, but massively more. Even if we build the AP1000 serially, do you think it could reach $3B/GW? That's a huge hike in reasoning, and I'm not sure where in the long long Wikipedia article for nuclear power it states that serially production of the same design could drop costs by a factor or 4 (or more).
Everywhere it has been studied, nuclear has a negative learning rate. See, for example, Figure 1 and the many articles it cites:
> You seem to be backing up my point with your claims. Nuclear has consistently gotten more expensive over time.
And we're also building fewer nuclear plants over time. Let's revisit your point:
> Even worse, nuclear very clearly gets more expensive the more if it we build, not less expensive.
This is not only untrue, it is the opposite of true. The more of it is built, the less expensive it is. The less of it is built, the more expensive it is.
You're right that nuclear has gotten more expensive. But that's because we're building less of it. If what you said were true - that nuclear get more expensive the more we build it - then today's nuclear plants should be cheaper than the ones built in the 1970s and 1980s during the nuclear boom.
> Even if we build the AP1000 serially, do you think it could reach $3B/GW? That's a huge hike in reasoning, and I'm not sure where in the long long Wikipedia article for nuclear power it states that serially production of the same design could drop costs by a factor or 4 (or more).
It's not a huge hike in reasoning. It's based on the price history of previous plant construction. Click on those plants built in the 1970s and early 80s. Many were built at a cost of only $2B. AP1000s are fundamentally not much different than previous PWR designs. Iterative improvements extend life and generate a bit more power, but the overall layout is the same.
In short we did build nuclear plants at scale, and it was 4-5x cheaper. This is a claim based on demonstrable patterns in price history.
"greater costs over time" wasn't what you claimed. What you wrote was:
> nuclear very clearly gets more expensive the more if it we build, not less expensive.
We've been building less nuclear not more. So if what you say is true, costs should have been going down over time. The reality is that nuclear is cheaper when built at scale. The larger number of plants being built together was cheaper than a handful of plants.
Your own source shows this. Look at this graph [1]. You see that cluster of plants around 1970? When we built a lot of nuclear, it was a lot less expensive. $1-2,000 per KW of capacity, or $1-2B per GW. This was actually cheaper than the figures I originally cited, thanks for the source I'll be sure to use it to demonstrate how nuclear is cheaper when built at scale in the future.
Likewise, Pripyat remains a ghost town because it was planned city built to support the power plant workforce. Now that the power plants are no longer in operation, there's no incentive to clean it up and resettle it.
South Korea's nuclear plants are also much cheaper, for the same reason.