Pareto efficiency is a micro-economics concept that seems most relevant to political considerations. The basic idea is, if one person can be made better off without anyone else being hurt, that action ought to be taken. Most systems are assumed to be at least in a local equilibrium of pareto efficiency, or at the very least, if they're not, economists are interested in why.
A simple example: Suppose a society has 30 units of labor to produce things. If they use some of them on guns, and some on butter there will trade offs between the two. More guns means less butter. However, if only 10 of the units are used, then the system will be Pareto inefficient. It's important to note that Pareto efficiency doesn't tell us anything about what allocation of guns/butter a society should choose, just that it makes sense to use all the resources available to you.
What does that mean in regards to writing code? Pareto improvements are software features/changes that improve the utility/enjoyment of your program for everyone using it. Fixing a bug is almost always an improvement. However, if the bug is now relied upon, then it is not an improvement to fix it. A common example are errors in writing APIs. Once released to the world and used by people, any changes to the API have costs, mainly, other folks have to rewrite their app. Adding in new features is not typically a Pareto improvement, because the cost of a menu item/adding in a hotkey may affect your users for whom the program worked just fine.
If it's easy/possible to make a Pareto improvement to your software you ought to. However, costs are often difficult to assess. Be carefully about assuming something is a Pareto improvement. However, once I'm sure it won't negatively affect users, I like to make the change. If it is not a Pareto improvement you have to weigh the competing concerns of all the people who will be affected the change. This is much more difficult to do, and often almost impossible to find the 'right' answer for.
Sunday, October 30, 2011
You know nothing Jon Snow
I finished reading the Dancing of the Dragons A Dance with Dragons. It was pretty good, and I'm amazed at how many stories/characters are in the ASOFAI. But, I just want to register a quick complaint. It's like 7 separate cliff hangers all arranged near the end of the book. C'est la vie I suppose.
Friday, October 16, 2009
The Black Swan -- Nassim Nicholas Taleb
I read this book a few months ago, but it resonated with me. In particular the book critiques using the bell curve and other standard statistical techniques as if they are infallible. Many issues in our world deal in exponential or extraordinary events. Our most recent economic crisis was a series of unlikely things all conspiring to create a recession. All of this leads one to consider black swans.
A black swan is a seemingly impossible thing that occurs and causes great ruin to those who bet against it. A simple example of this is if a bank is making loans to farmers and only farmers. While it is unlikely that the bank won't make money, if all the topsoil were to simply blow away the bank would almost certainly fail. A good black swan is making it as a muscian, making it as a reality TV star, and getting on the NY Times bestseller list.
So, things to take away (in the style of the book)
Is this a measure that lives in mediocrastan?
Does this thing benefit from Black Swans?
Can one maintain a bleed strategy? (Many small losses, few HUGE wins)
Absence of proof =/=> proof of absence
Scaleability => fractals and similarity (NOT normality)
hyper-conservative "risk-taking" (reduce exposure to negative black swans)
hyper-aggressive in areas "needing caution" (Maximizing exposure to positive black swans)
skeptical about confirmations
no-nonsense/practical about academic matters
intellectual about practical matters
"Don't run for trains" It is more difficult to lose at games you set up.
Using a measure (profits) as a cause for survival is circular. (Post-ante selection bias is bad!)
Imagine the million failed people. 'I'd rather be lucky than good'
A Random Walk Down Wall Street - Malkiel
I'm working on reading through this book. It's a book about the financial markets written with the intended audience of investors. It has been recommenced to me off-handedly by my Money and Banking professor from spring '09. I'm a third to a fourth of the way through.
It's funny to hear descriptions of quants being able to value a firm with, 'a few presses of a calculator'. Now-a-days valuation of firms seems to be done in time measurements where electron cycles are a relevant time division.
The best part of reading this book is the history it provides. The author is attempting to explain contemporary methods for picking stocks. But because this 4th edition book was published in 1985 it's now an explanation of history. Understanding the Firm Foundations theory and Castles in the Sky theory helps to contextualize the criticism of economists and other thinkers of speculation in the stock market today. The "Black Swan" gains additional meaning now that I understand the object of critique as well as the critique.
(MORE TO COME...)
Subscribe to:
Comments (Atom)