If the problem is that we don't have a lot of subject cataloging, provide more subject cataloging. We could copy the LoC cataloging for most of the catalog without too much work.
If the problem is literacy instruction, then we should work on a list of books for literacy, not rely on some tool that can't tell the difference between a 17th century children's book and a 20th century one, or how much dialect is used.
While I agree that it would not be worth adding readability score if it had much impact on these and other worthy goals, I really don't see it as either/or. Granting of course that adding scores will take some time away from other projects (and, that it's not my personal time at stake here), I still see this as relatively high gain for relatively low investment. There are lots and lots of cool things that could be done with the catalog. And, any relatively "easy" (i.e. automated) method of adding readability scores will inevitably miscategorize a whole bunch of books. But, I think the 'signal' will far outweigh the 'noise'. Even in the context of the above, the scores would provide a great starting point for being improved with manual cataloging and literacy labeling. Don't let the perfect stand in the way of the good. Plus, I think the scores (and miscategorizations) are interesting in and of themselves for those of us interested in words and language. Cheers, Scott S. Lawton http://blogsearch.com/ - a starting point http://ProductArchitect.com/ - consulting