First of all you need to use a good resize algorithm for that case. Lanczos, or Photoshop's Bicucbic optimized for reduction. And then to make for better contrast around the letters you can use some output sharpening.

Secondly, there is a minimum resolution you can use to render the fine letters. The book is a perfect example because of the different fonts and sizes.

Below is bicubic on the left, lanczos on the right, in two different sizes (10% and 20%). You see that lanczos separates letters like I better than bicubic. You can read the top line in the smallest image with lanczos but not with bicubic. Also twentieth edition is better rendered with lanczos. Also the 2013 is prettier rendered, albeit readable.

You need to make the image bigger to read the bottom lines. Again, lanczos renders I's with more "Air" to the letters around them. The differences are subtle, but may have big impact in the long run if a person really wants to read more text like this.

Lanczos also allows you to resize books and pages with fine texture with less chance of causing fake texture (moire).

A 3rd thing to note is that the results I posted are based on a jpeg compressed source, which I then saved as jpeg again, and maybe S.E. recompressed it again. If you keep everything in raw (tiff/bmp/ppm uncompressed) until the last moment and then use a subtle compression (maybe even lossless) you will also improve the readability.