How good are doctors?

Atul Gawande’s The Bell Curve in last week’s New Yorker

It used to be assumed that differences among hospitals or doctors in a particular specialty were generally insignificant. If you plotted a graph showing the results of all the centers treating cystic fibrosis—or any other disease for that matter—people expected that the curve would look something like a shark fin, with most places clustered around the very best outcomes. But the evidence has begun to indicate otherwise. What you tend to find is a bell curve: a handful of teams with disturbingly poor outcomes for their patients, a handful with remarkably good results, and a great, undistinguished middle.

It’s an excellent look at how honestly hospitals are dealing with their patients. Also of note is Tom Peter’s reaction.

Ethnoclassification and retrieval

I’ve been doing some product development consulting for a content-centric company, and started to wonder if ethnoclassification could benefit them somehow. I was hard-pressed to think of how, after the content was tagged, the tags would lead to better sets of content. Although they have a lot of content, they only have a few authors who do the classification. When I returned to what others have written on the topic, I think the idea of ethnoclassification was so enthralling we didn’t look as closely at how it facilitates retrieval and the relationship to the content set and authoring process.

It seems that selecting a classification technique should factor in how much control we want over retrieval, how much content there is, and how many authors there are. Ethnoclassification becomes more useful when there’s a ton of authors and you can start to see relationships emerge among tags.

For example:

  • Large content set, many authors, low control over retrieval: Ethnoclassification makes sense when seredipitous retrieval is desired and a critical mass of authors are generating tags, like on Flickr
  • Large content set, many authors, high control over retrieval: A robust structure is needed to ensure end users can find information, like the ontology-driven
  • Medium content set, medium authors, medium control over retrieval: This situation allows for interesting hybrids, like the Pratt Talent site that balances ethnoclassification with a controlled vocabulary
  • Medium content set, medium authors, high control over retrieval: traditional controlled vocabularies/taxonomies make sense, and scale down to the point where there is so little content/few authors that no classification is needed

Sorry, I’m not sorry

I’ll go on the record here and say that Sorry Everybody is contrary to the ideals that make America what it is. I voted against Bush, but that doesn’t make me want to apologize to the rest of the world that he won. He won because we are a Republic with a democratic voting process that elected him. To say that we screwed up just because our guy didn’t win is to say there’s something wrong with democracy. Until something better comes along, I’ll stick with democracy, thanks.

Distributed processing on a chip

Details are emerging on IBM’s “supercomputer on a chip”, which essentially seems to take the logic that distributes operations to multiple chips that used to be done in applications or the operating system and integrate it at the chip level. This has the potential to exponentially speed up everything that’s not already a supercomputer, as software doesn’t have to change to accommodate more than one processor per machine.

From a design perspective, we could think of this as a process innovation rather than one of hardware engineering, as the advantage was gained by taking work from one stage of the system and moving it to another stage.