Love it. Just wanted to share this book (which I think we discussed last year?) called "Technopoly: The Surrender of Culture to Technology" by Neil Postman [link](https://en.wikipedia.org/wiki/Technopoly) which I really liked. If someone likes this post then they would love this book (if they haven't already read it!). It's from the early 90s -- very prophetic.
The post and the title of the Postman book made me think of a podcast (EconTalk) where Russ Roberts interviewed Jerry Muller about his book, "The Tyranny of Metrics". (https://a.co/d/5oqUzE6).. I did not get around to reading the book but feels very related. I am a slow reader too :(
"Technocracy applies a veneer of science used to justify the moral positions of those in power."
This may very well go into my file cabinet of pithy wisdom to pull out when needed. I can put it in the same drawer as Sowell, "There are no solutions, only tradeoffs."
Good rant. However, I want to push back on your point about COMPAS. A perfectly calibrated risk prediction can still be extremely lacking. I find it hard to understand why risk scores far from 0 or 1 are useful when it comes to decision-making. When an outcome is highly variable given observed attributes, shouldn't that tell us that we need to collect more data, or focus our technocratic energy elsewhere? For COMPAS specifically, similarly accurate predictions can be achieved by logistic regression on two features (age and number of prior convictions): https://www.science.org/doi/full/10.1126/sciadv.aao5580
To clarify, I am not arguing in favor calibration! And I'm definitely not in favor of COMPAS. I was trying to use that example to show calibrated models can still be unfair. I should edit the post to highlight your point about how it's not better than that two feature predictor anyway.
A freely swinging pendulum eventually settles into stillness if not re-stimulated. Balance (fairness) comes over a long period of time rather than immediately b/c we have discovered how to jump from wide swings of bias to immediate fairness. Mathematics does not seem to me to be the answer to questions of fixing racism, sexism, and other "isms" of the past. Temporary swings in the other directions create balance in the long run...perhaps that long term view is the only real solution. Chaos Theory and fundamental fractals are at work here, in my opinion, especially when we transfer methods (tools) from one problem to another.
"People who study policy mean well, but they convince themselves that because they spend so much time on it, they know better than everyone else. They then want to leverage their elite status to impose their ideas on everyone else. Technocracy applies a veneer of science used to justify the moral positions of those in power. "
Sometimes I wonder if some of the problems people feel with longtermist/utilitarian movements have a at least a somewhat similar flavor. Maybe it would fit a little better if you replaced "study policy" with "study philosophy, math, and cs".
Newsflash: the world is unfair. The individual handles the real world by doing what is right regardless of there being fair rewards or not. (External) fairness is irrelevant for good character. What would the Stoics think about this algorithmic nonsense? You control the effort not the outcome. But then... is there even free will?
Love it. Just wanted to share this book (which I think we discussed last year?) called "Technopoly: The Surrender of Culture to Technology" by Neil Postman [link](https://en.wikipedia.org/wiki/Technopoly) which I really liked. If someone likes this post then they would love this book (if they haven't already read it!). It's from the early 90s -- very prophetic.
It's on my list! But man, I am such a slow reader.
I am also a slow reader! But, I pinky promise that it's worth it :)
The post and the title of the Postman book made me think of a podcast (EconTalk) where Russ Roberts interviewed Jerry Muller about his book, "The Tyranny of Metrics". (https://a.co/d/5oqUzE6).. I did not get around to reading the book but feels very related. I am a slow reader too :(
I have read that book a while ago, and it is (a) very related, (b) very good.
..and my reading list continues to grow...thanks, I will get this and read it.
"Technocracy applies a veneer of science used to justify the moral positions of those in power."
This may very well go into my file cabinet of pithy wisdom to pull out when needed. I can put it in the same drawer as Sowell, "There are no solutions, only tradeoffs."
Good rant. However, I want to push back on your point about COMPAS. A perfectly calibrated risk prediction can still be extremely lacking. I find it hard to understand why risk scores far from 0 or 1 are useful when it comes to decision-making. When an outcome is highly variable given observed attributes, shouldn't that tell us that we need to collect more data, or focus our technocratic energy elsewhere? For COMPAS specifically, similarly accurate predictions can be achieved by logistic regression on two features (age and number of prior convictions): https://www.science.org/doi/full/10.1126/sciadv.aao5580
To clarify, I am not arguing in favor calibration! And I'm definitely not in favor of COMPAS. I was trying to use that example to show calibrated models can still be unfair. I should edit the post to highlight your point about how it's not better than that two feature predictor anyway.
All of these complex risk scores are always blurry jpegs of conventional wisdom. I'm against all of them, as I wrote here: https://argmin.substack.com/p/healthcare-and-the-ai-dystopia
A freely swinging pendulum eventually settles into stillness if not re-stimulated. Balance (fairness) comes over a long period of time rather than immediately b/c we have discovered how to jump from wide swings of bias to immediate fairness. Mathematics does not seem to me to be the answer to questions of fixing racism, sexism, and other "isms" of the past. Temporary swings in the other directions create balance in the long run...perhaps that long term view is the only real solution. Chaos Theory and fundamental fractals are at work here, in my opinion, especially when we transfer methods (tools) from one problem to another.
"People who study policy mean well, but they convince themselves that because they spend so much time on it, they know better than everyone else. They then want to leverage their elite status to impose their ideas on everyone else. Technocracy applies a veneer of science used to justify the moral positions of those in power. "
Sometimes I wonder if some of the problems people feel with longtermist/utilitarian movements have a at least a somewhat similar flavor. Maybe it would fit a little better if you replaced "study policy" with "study philosophy, math, and cs".
Newsflash: the world is unfair. The individual handles the real world by doing what is right regardless of there being fair rewards or not. (External) fairness is irrelevant for good character. What would the Stoics think about this algorithmic nonsense? You control the effort not the outcome. But then... is there even free will?
Is the 'world' unfair or is it dynamic?
Social problems do not have mathematical or algorithmic solutions.
Technocrats & transhumanists disagree with you, sometimes violently....