10 Pillars of Knowledge: Mapping the Knowledge Covered by Library Classification Systems

Michael Heaney
Executive Secretary, the Bodleian Libraries
Clarendon Building, Bodleian Library, Oxford UK

Chaim Zins: We did not quantify the differences among the 4 systems. This was a qualitative comparative analysis. It was not easy.

Michael Heaney: But what concerns me more is the imposition, on top of the assessment of difference, of an evaluative layer which presumes a priori that 10PK is superior to the others. (May 9, 2011)

Chaim Zins: We selected 10PK as an evaluative model for the study since, to the best of our knowledge, it is a systematic and comprehensive map of contemporary human knowledge. It meets the three conditions of the ideal knowledge map:

The ideal knowledge map should meet three conditions. First, its categories are mutually exclusive, meaning they do not overlap. Second, the categories are collectively exhaustive, meaning that together they compose all the relevant categories. Third, the map can represent all the relevant items without exception. This means that the ideal map of human knowledge can represent all fields of knowledge; that is, every field of knowledge belongs to at least one category of the map. (from my coming book)

As we proved in the paper LCC, DDC, and UDC fail to meet the three conditions.

Michael Heaney: Relativistically, one could equally say that 10PK fails to match up to the standard of UDC (or whichever you choose). (May 9, 2011)

Chaim Zins: No. An ideal map of human knowledge needs to meet the 3 conditions. Since UDC fails to meet the 3 conditions it should be revised.The same with LCC, and DDC.

Michael Heaney: As an example of the potential lack of rigour in 10PK, its analysis of ‘Religions’ seems to me to be very ad hoc:
3.1 Ancient
3.2 Monotheism (sc. Monotheistic)
3.3 Asian
3.4 Modern
3.5 Ethnic

This mixes geographic, chronological and typological analysis.
Is Mormonism ‘Monotheistic’ or ‘Modern’? And whatever the answer, why should
that be so? (May 9, 2011)

Chaim Zins: The mapping of the supernatural phenomena was not easy. Scholars in religious studies disagree on the definition of "religion" and on the classification of world religions (I related to these issues in my coming book).

The classification of the world religions (cat. 2.3) is based on an empirical study. Consequently the 5 categories (2.3.1-5) are by definition a taxonomy (not a typology) of world religions (what you call "ad hoc"). Note that taxonomies are usually ad hoc since they are usually based on ad hoc collections of the classified items.

Even if you are right and the ad hoc taxonomy (cat. 2.3.1-5) is incomplete and unsystematic it would not refute our findings and conclusions regarding LCC, DDC, and UDC.

Certainly, it will show that one of the 10PK's third level hierarchical classifications (out of 9) needs to be revised. It will not prove that 10PK's first level classification (i.e. the 10 pillars), and the second level classification (i.e., cat. 2.1-2.3) are problematic.

Still, I claim that the classification of world religions (cat. 2.3.1-5) meets the 3 conditions. The 5 categories represent 5 characteristics of religions rather than 5 different models for classifying religions. The misinterpretation is understandable and lies in the ambiguity of the terminology.

I identified 5 main groups of religions. The terminology is aimed to represent the main characteristics of these groups:

* "Ancient" = religions that were practiced in the past (and not to religions of the ancient era).
* "Modern" = new religions that were recently created (not religions of the modern era)
* "Asian" = religions that were originated in Asia, and are practiced mainly in Asia (not to religions practiced in Asia).
* "Monotheism" = the 3 historic monotheistic religions (Judaism, Christianity, and Islam) Note that I first used the adjective "monotheistic", but 2 English editors suggested using the noun. What do you suggest?
* "Ethnic" - This is a descriptive (as opposed to evaluative) politically correct terminology. Do you have a better PC suggestion?

Your analysis of the 5 terms is highly important. Thanks. I will consider revising the terminology in the next version of the map.

Michael Heaney: Where does the Mayan religion fit? And who among us is not ‘Ethnic’?" (May 9, 2011)

Chaim Zins: The Mayan religion is not part of the 10PK map. It is being mapped by the 10PK map. As long as you can place it in at least one of the 5 categories the 10PK's taxonomy of world religions (i.e. cat. 2.3.1-5) meets the third condition. The same goes with any of the hundreds (or thousands) world religions.

May I refer you to the theoretical foundations of the 10PK project.

Michael Heaney: The only basis for promoting 10PK ahead of the others is that it CLAIMS to be a theory of knowledge while the others only CLAIM to be library classification schemes (based more or less explicitly on literary warrant). (May 9, 2011)

Chaim Zins: 10PK is a map of human knowledge. It can be developed into a library classification scheme. LCC, DDC, and UDC are "officially" aimed at classifying bibliographic items. Nevertheless, I adopt the broad definition of "knowledge map"; accordingly 10PK, LCC, DDC, and UDC are knowledge maps.

Michael Heaney: But the mere claim is not enough to justify its promotion as an absolute measure, let alone to then add value judgements to the assessment (‘they failed to adequately and systematically present contemporary human knowledge’). (May 9, 2011)

Chaim Zins: 10PK has solid theoretical foundations (see bellow). We justifiably selected it as an evaluative model. Based on this methodological decision we based the comparative analysis. By no means we do not claim that 10PK is the only possible model for evaluating LCC, DDC, and UDC. Still, our findings and conclusions are valid.

Michael Heaney: I would advise looking at the work of the Classification Research Group and people like Barbara Kyle back in the 1960s, and/or its grandchild the revised Bliss Classification Scheme.

The theoretical weaknesses of schemes are well acknowledged but are rooted in their historical development. It will be interesting to see how 10PK fares over a hundred-year timescale. (May 9, 2011)

Chaim Zins: I definitely agree with you. (May 11, 2011)

Michael Heaney: Thank you for this.

My only further comment is that my remarks about the classification of religion were intended to demonstrate that 10PK does not meet these two requirements:

categories are mutually exclusive, meaning they do not overlap. (Mormonism is both modern and monotheistic)
categories are collectively exhaustive, meaning that together they compose all the relevant categories (Mayan is neither Ancient, Asian or ‘Ethnic’ (where ‘Ethnic’ seems to be defined, by example, as primitive/animistic)

I could add moderm Paganism, which is demonstrably Modern (we have the documentary evidence) but is also ‘Ethnic’ in that it is primitive/animistic. (May 11, 2011)

"I wonder how did you come to the conclusion that you can compare subject coverage of any knowledge field by taking three top levels of three biggest aspect classifications systems (the smallest of which Dewey has 40,000 classes, the biggest has 300,000 classes and the one of them which has 68,000 is completely synthetic and creates subjects by combinations of concepts)." (March 2011)

Chaim Zins: This is a very important argument. We explicitly responded to this argument in the article (see methodological considerations):

"Scholars who are familiar with the three systems have raised two arguments about the methodology of the study and, consequently, its findings and conclusions.

The first argument questions the focus on the top levels of the hierarchical structures. The library schemes may not address the topics outlined in the 10 Pillars of Knowledge at the top levels of each scheme, but nearly all are covered by deeper levels of the schemes.

In fact, at the beginning of the study, we delved into the hierarchical structures and mapped the relevant topics wherever they appeared in each library scheme. However, we decided to leave this practice and focus on the first two levels of the hierarchical structures. We realized that if we go all the way down the hierarchical structures, then each of the systems covers all the topics. Because LCC, DDC, and UDC are aimed at classifying publications, they have to provide a practical solution for every publication. Practically, this means that the three systems cover all topics and all fields. However, very often the practical solutions seem strange. For example, as noted above, veterinary medicine is presented in LCC by the SF (animal culture) subclass, which is a subclass of S (agriculture). Obviously, veterinary medicine is not a subfield of agriculture. Apparently, librarians and users need practical guides to LCC, DDC, and UDC. But developing such guides was not the goal of this study.

A hierarchical classification should meet three conditions. First, the categories are mutually exclusive, meaning they do not overlap one another. Second, the categories are collectively exhaustive, meaning together they comprise all the relevant categories. Third, the classification can represent all the relevant items without exception. This means that every field of knowledge is represented (i.e., can be placed) by at least one category. LCC, DDC, and UDC may meet the third condition and cover all fields of knowledge, but do they meet the first two conditions?

We decided to focus on the first two levels of the hierarchical structures because these levels embody the essence of the classification system. Each level is viewed as an independent classification unit and needs to meet the three conditions. The main classes of each classification (i.e., the first level) need to be mutually exclusive, meaning they do not overlap one another, collectively exhaustive, meaning together they comprise all the relevant classes, and represent all the relevant fields without exception. This means that every field of knowledge is represented by at least one of the main classes.

Each of the main classes establishes an independent classification unit, which comprises main subclasses (i.e., the second level). The main subclasses need to be mutually exclusive, collectively exhaustive, and represent all the relevant fields without exception. This means that every relevant field is represented by at least one of the main subclasses.

In retrospect, this theoretical-based, methodological decision to focus on the top levels of the hierarchical structures is justified by the findings of the study. The top levels, in many cases, are not systematic.

The second argument questions the focus on the terminology of the hierarchical tree (i.e., classes and subclasses) while ignoring auxiliary tables and application guides. The library systems include auxiliary tables and guiding tools that may refer to all fields of knowledge. Ignoring them affects the findings and conclusions of the study.

In fact, at the beginning of the study, we related to auxiliary tables and tools, but we decided to leave this practice and focus on the terminology of the hierarchical structures. A classification system is a structured network of concepts. The structure is essential as it represents the logical relations among the linked concepts. The terminology is a fundamental element and comprises the structure’s building blocks. It needs to be self-explanatory: “What you see is what you get.” The auxiliary tools are not part of the structure, but rather explain it. In retrospect, this theoretical-based, methodological decision to focus on the terminology of the systems is justified by the findings of the study. The terminology, in several cases, is biased and misleading."

An anonymous responder: I found literature supporting this article unacceptably poor. (March 2011)

Chaim Zins: The findings and the conclusions of the article are based on a thorough and critical analysis that took us more than two years to accomplish. They are not based on sheer literature review.

We could list literature that support our conclusions but we do not think that it is relevant for substantiating the findings and conclusions.

While conducting the study we checked hundreds of references but we decided not to list them for three main reasons:

(1) We did not find any particular reference that systematically and comprehensively critically analyzed and compared the three systems. In this respect our study is unique.

(2) We did not find any particular reference that validate or weaken our findings that are exclusively based on critical analysis.

(3) As a matter of principle, we do not cite any reference that we can not check with our eyes. If by no means you find relevant references please send me a copy and I will respond.

Finally, in our paper we could not list all of our findings. Our conclusions are solid. Our paper is important because we have succeeded to systematically validate what scholars in the field of knowledge organization know for years.