« The house price bounce is (mostly) real - for now | Main | The "Cripes!" growth model »


Feed You can follow this conversation by subscribing to the comment feed for this post.


Why not just blog about some of the things you're called upon to think about as associate dean?

You might be interested in the computer science equivalent of prioritization of processes on a processor when doing concurrent processing. Switching between processes (called context switching) is very costly in time, as it involves saving the state of the processor to memory, loading the next process, etc. And in the case of using disk memory, which is very slow, it's akin to working on a project, with all your papers spread out on your desk. Before you finish, you need to work on another project, so you gather the papers into a folder and put them into your filing cabinet. Only, your filing cabinet is on the other side of the continent--from where you will need to retrieve the files before you can resume. And yet, the illusion of parallelism that is achieved through concurrency is worth this overhead.

This overhead reduces the amount of time the processor can spend thinking in aggregate, but increases the amount of things it can be thinking about 'at once'. But the more things want it to think about at simultaneously/in a given time frame, the less actual thinking it can accomplish.

DW: Hmmm. I might just do that, and see if there's any interest. Some of it actually has economic application. Like the violation of Walras' Law in the Barro-Grossman repressed inflation regime, when we had an excess demand for seats in courses back a few years ago. (A student might only want one more course, but would register an excess demand for 5 courses, trying 4 more when he couldn't get into the first one). But some of it's a bit too confidential/sensitive.

Andrew: I hadn't thought of switching costs (from one topic to another). I had been thinking about fixed costs of just getting started on thinking about a topic. If timing doesn't matter, you would never switch; you would always finish one batch of thought before starting on the next, to avoid any switching costs. But then the phone rings, someone comes by your office, or there's a meeting. So timing matters, and so those switching costs matter. You are right. Good model. One more thing to think about.

But there are times when it is a waste to continue thinking on a topic at that given time. Maybe you are waiting for more information, etc. Processes will tell the kernel (the boss process that hands out the resources) that it's done what it can for now, and rather than tie up the processor. This is done when it is passively waiting for an event, such as input (a mouse click, or a disk access, etc.) to occur.

I suppose a think-tank would be like an open system, where different units are given topics to work through, hopefully each performing their assigned thought process the most efficiently. Then, if one were to add another person to the mix, it would be like opening trade with another country, and topics might rearrange appropriately. Would there be a period of decreased output while tasks are redistributed? Or would the process be gradual(perhaps waiting for resolutions to the current problems before transferring similar problems to the new guy, similar to contracts running their course before getting outsourced)?

From a macroeconomic point of view, I would suggest focusing on too much cheap labor, too much cheap debt, asset prices being too high, too much wealth/income inequality, and a "fungible" money supply that is skewed towards too much debt.

The comments to this entry are closed.

Search this site

  • Google

Blog powered by Typepad