So a few people have said that books chosen can't possibly be ordered and that it's pure opinion which ones are better and better for my purpose than others. Given that any set of objects can be placed in order by means of an ordering function, below is my proposed ordering function. Any set of books that appropriately maximizes or minimizes these criteria compared to any other set of books will be most preferred.

  • Readability to a broader audience though this won't be the general public. This is intended to avoid problems like Principia that's basically unreadable to someone who wasn't a personal friend of Newton's (okay, I'm probably exaggerating but you get the idea).
  • ‎Breadth of Coverage across all three books.
  • ‎Content of each book should be strongly connected to the other two. There will be differences in terminology that arise as concepts not described in the books are rediscovered by the future scientists (they will come up with different names for things) . Having one book lead to another minimizes this kind of problem.
  • ‎First, or first clearest expository of a foundational idea/concept/system for that field. (Principia is first but someone else wrote a clearer explanation later. Thus, the latter work is preferred over Principia. )
  • ‎Recognizable to someone in that field, at the target entry time, that the material in the book is about that field. (someone working with the newly rediscovered Maxwell's equations isn't going to be able to jump to semiconductors or fiber optics. Well, I'm guessing they won't be able to.)
  • ‎Corrects misinformation or model failures in knowledge in this field at the target entry time. (Pick your favorite "I can't believe the early people got that wrong" moment. Fixing those moments is important and the primary hope of these books. )

So my question is: with the above constraints on the books possible for the Three Books questions, would the questions be sufficiently narrow and not opinion-based?

  • 2
    $\begingroup$ So... what's your question? $\endgroup$ – a CVn Jan 15 '18 at 9:28

No way.

Let's have a look at the numbers of books listed by the US Library of Congress. Number of books by keyword (and some subfields) as of Jan 25, 2018:

  1. Astronomy: 14,444
  2. Chemistry: 43,725
  3. Medicine: 82,493
  4. Physics 37,529

It is true that some of these books may constitute catalogs of sort, but then, why not a catalog of chemical compounds to share all the available synthetic objects that we have so far managed to create?

I honestly have a hard time imagining someone who would be so familiar with all these texts in any of these disciplines in order to provide a reasonable justification of why their specific subset is more fitting than most other subsets (and I am not even thinking of the best subset).

On the other hand, one could narrow down to some very specific sub-discipline, where only a handful of books exist. The list of human disciplines is quite large. I imagine that WB would restrict the number of three books questions to only a subset of these. Then, again, I have a hard time imagining why only a subset of these should be allowed, and the rest discarded.

If you want to further play around with the numbers, you can construct the link from


  • $\begingroup$ On the other hand, one could construct a machine learning algorithm to parse the text of all these books, run the scoring system proposed by Green, and provide an ark of human knowledge for posterity... at least for the few thousand that are available online. Perhaps it is a question better fitting for codegolf.stackexchange.com ;) $\endgroup$ – NofP Jan 25 '18 at 14:29

This will probably work for rather low-level natural sciences but not for social sciences or high-level theoretical natural sciences. Your criteria are based on several assumptions:

  • there is one and only one 'correct' and unified theory in the field;
  • the entire field uses the same terminology to mean exactly the same things;
  • the concepts/main points of this theory can be summarised in a brief concise manner.

This is not the case in social sciences. Psychology, for example, does not have one dominant school or theory. Moreover, theoretical psychology (research) and applied psychology (therapy) operate quite differently. Psychologists also have a habit of 'borrowing' terms and giving them slightly different meanings, hence, any paper or book has a long part with definitions. Moreover, these definitions may change over time as a theory develops, as was the case with Freud and psychoanalysis. The same can be observed in other fields of humanities as well. It does not mean that they are somehow deficient. But they are just different from natural sciences in their organisation. So you need to come up with a special approach to them.

I also wonder about technical and engineering manuals. You probably want those to rebuild technology. Do they need theory whatsoever? Perhaps a book of blueprints and accompanying manuals on ore refining/tool making/woodworking/etc. would be a better fit.

I would suggest abandoning the one-fits-all approach and narrow down your field of interest first or divide fields into groups if you are looking into building a comprehensive library. Then you can come up with fitting criteria for each group of fields.


You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .