ISKO Italy Open conference systems, Paradigms and conceptual systems in KO

[Seminar] Epistemological paradigms in KO

Birger Hjørland

Building: Main building
Room: Hall VI
Date: 2010-02-25 02:00 PM – 04:00 PM
Last modified: 2010-03-01


1.    What is epistemology?

Epistemology is also termed “the theory of knowledge”. Any kind of scientific or scholarly methodology is related to epistemology, as is everyday thinking.  We are all influenced by epistemological theories, whether we realize it or not. Epistemology is the core in the question “what is the scientific method?”  (or  “what are the scientific methods?”).   Epistemology is essentially about methodological ideals and commitments.

Epistemology is often defined as “a branch of philosophy”. I have two problems with this 1) the metaphor of a tree a model of knowledge organization 2) The isolation of the field to one academic discipline. In my opinion the network is a better metaphor of KO and epistemology is informed not just by professional philosophers but also by scientists and by several metafields like sociology, linguistics and psychology (and the broader field of science studies).


2.    Epistemological positions (or theories)

There are many different views or positions that are used or discussed in different literatures today. Among them are:

Activity theory & The socio-cultural perspective

Actor-network theory (ANT)

Authoritanism, dogmatism, opportunism etc.

Common sense



Critical rationalism (Karl R. Popper)

Critical theory



Epistemological anarchism (Paul Feyerabend)

Feminist epistemology & Standpoint epistemology



Logical empiricism

Marxist philosophy of science

Operative constructivism (Niklas Luhmann)

Paradigm-theory (Th. Kuhn)







Semiotics & Social semiotics


Social constructivism

Social epistemology

Socio-cognitive view

Structuralism and post-structuralism

Systems theory

Each of these positions may have some proponents or use in LIS and KO. However, even if some people say that they are using a specific approach (say empiricism, pragmatism, feminist theory or social constructivism) it is not given that they understand this position the same way as other researchers using the same label. To find out exactly what kind of methodological commitments that should be associated with a given label is in itself a serious scholarly job that has to be done.

I edit “a lifeboat” as a tool to help you identifying core characteristics in the different positions that may be encountered in your studies.



In my opinion all these different positions have some basic problems in common such as the relative roles of observation and thinking, whether theories and interests influence observations, the role of language, culture and community in knowledge production, knowledge organization and –use. By considering such basic issues, the core epistemological positions may turn out to be:

·        Empiricism

·        Rationalism

·        Historicism

·        Pragmatism



“Empiricism” should not be mixed up with “empirical research”. It is today generally accepted that research and science should somehow be empirical. The four mentioned approaches should be regarded as different ideals for doing empirical research (although this may at first seems strange, in particular about rationalism). If this was not the case other views would simply be obsolete and not worth considering outside the history of philosophy.

Empiricism is the ideal of basing knowledge on observations (and on inductions from a pool of observations). Theoretical selections and interpretations of observations have to be avoided. (If not, we have moved to another approach). Observations are seen as “given”, not as contextual or theory-dependent. Empiricist approaches are based on bottom-up analysis (information is read out of objects, perception is seen as a receiving or a passive process).  It is thus typical of empiricism’s ideal and methodological commitment that it only relies on features that can be observed and tries to avoid theoretical selection of which properties are most important because such theoretical considerations are considered problematic from this methodological ideal.



Rationalism is the ideal of basing knowledge on logics, principles, rules and idealized models.  Rationalism is skeptical about empiricism and about sense experiences that are not organized according to principles, which are—in one way or another — a priory in relation to experience (for example fast-wired into our cognitive systems). Rationalism may seem strange from a modern, empirical point of view. It is however a strong classical position with arguments which cannot be ignored:  Every process of obtaining information is always an interaction of bottom-up and top-down processing where rationalism is needed to account for the latter kind of information processing. Reality is interpreted in logical concepts and categories. “Logical division” is a rationalist method (and in knowledge organization the facet-analytic approach comes closest to this ideal). Empirical research inspired by theoretical models should be considered rationalist if these models are not themselves derived empirically.



Historicism is the ideal of basing research on social contexts, on historical developments and on the explication of researchers’ pre-understanding. It is based on the understanding that observations are “theory-laden”, or culturally influenced (as opposed to neutral and “objective”) processes. Hermeneutics belongs to historicism as I understand it. In retrospect Kuhn's book (1962) Structure may be seen as an hermeneutic interpretation of the sciences because it conceives scientists as governed by assumptions, which are historically embedded and linguistically mediated activities organized around paradigms that direct the conceptualization and investigation of their studies



Pragmatism is the ideal of basing knowledge on the analysis of goals, purposes, values and consequences. It is a kind of Darwinism applied to epistemology (knowledge is understood as developments made in order to increase humankind’s adaptation to the physical, biological and cultural environment). It is closely related to historicism by understanding that observations are contextual and “theory-laden”. The difference is that pragmatism tries to be explicit about the purpose of research and cognition. 

The basic methodological commitments in these four positions are related to what is considered relevant information:


 Simplified relevance criteria in four epistemological  schools

(from Hjørland, 2002, p. 269)





Relevant: Observations, sense-data.  Induction from collections of observational  data. Intersubjectively controlled data.




Non-relevant:  Speculations, knowledge transmitted from authorities. "Book knowledge" ("reading nature, not books"). Data about the observers' assumptions and pre-understanding.


Relevant: Pure thinking, logic, mathematical models, computer modeling, systems of axioms, definitions and theorems.




Low priority is given to empirical data because such data must be organized in accordance with principles which cannot come from experience.

Relevant: Background knowledge about pre-understanding, theories, conceptions, contexts , historical developments and evolutionary perspectives.


Low priority is given to decontextualized data of which the meanings cannot be interpreted. Intersubjectively controlled data are often seen as trivia.


Relevant: Information about goals and values and consequences both involving the researcher and the object of research (subject and object).




Low priority (or outright suspicion) is given to claimed value free or neutral information. For example,  feminist epistemology is suspicious about the neutrality of information produced in a male dominated society.


It should be emphasized that epistemological theories are competing views. One view is not as fruitful as any other view. The task is to work out a fruitful epistemological position. In my opinion historicism/pragmatism is a better choice compared to empiricism, rationalism and logical positivism. If other people are not convinced we need a dialog and further work to reach an agreement. The current fragmental state of purely constructed views and little coherence in our field is not the best position for a field.




3.    The double role of epistemology in LIS and KO

LIS/KO is a field of study depending on methodological commitments and thus an epistemological position.

On the other hand is LIS/KO studying other fields which may be seen – more or less - as a mixture of competing epistemological view. In order to organize knowledge about (say) the arts, one need to know about the different competing views of art (which again are related to emistemology).


4.    Ontological, epistemological and social dimensions of knowledge organization

Distinctions have been made between “the intellectual” and “the social organization” of knowledge (Whitley, 1984), or – as discussed by Hjørland & Hartel (2003) -  “ontological, epistemological and sociological dimensions of domains”


Ontology is about what exists in the world. Bunge (1977) in a book about ontology use the metaphor that it is about “the furniture of the world”.  The idea to classify according to, for example, “the theory of integrative levels” (from subatomic particles to planetary bodies and stars, from single cells to particular organisms and particular societies, and so on) is an example of an ontological classification.


To divide people into social groups (e.g., librarians, computer scientists, biologists and lawyers) and to organize knowledge in relation to such groups is by contrast a social organization of knowledge. Most research, education, publications, conferences etc are organized according to disciplines or professional forums. The disciplinary structures of journals as well as cooperation between scientists (and thus citation patterns) are thus examples of social organization of knowledge.


The epistemological dimension is related to the way things are being investigated. Because a given approach makes certain things visible and others relatively invisible epistemologies have implications for what is being described . There is thus a dialectic influence between ontological and epistemological views.  For example, when behaviorism became a view in American psychology (related to positivist epistemology), only observable behavior could be reported. A concept like “memory” became replaced by “delayed response” because only the last phenomena can be observed directly. Behaviorism and positivism thus imply their own ontology. Let us consider an example from LIS itself: If we subscribe to the bibliometric approach to knowledge organization, then certain kinds of entities are in the focus and are organized (e.g. authors, journals, disciplines), while other things are being neglected (perhaps genres, species and facets).  Thus, different epistemological views, such as empiricism, rationalism, historicism and pragmatism may not turn out to describe the world in the same way: They tend to bring with them their own basic sets of furniture.


This also means that any ontological claim has to be justified by empirical, methodological and theoretical arguments.  If anybody claims to describe what exists in the world, we should ask: From what perspective do you see that? And with which methods have you observed it? It is not possible to observe from nowhere and believe that this is the real furniture.  The American philosopher Willard Van Orman Quine found that the study of ontology should be based by drawing conclusions from the theories of the, especially from their canonical formalizations. He found that scientific theories are extensions of the theories we develop and use informally in everyday life, but they are our best sources of knowledge as to what the world is like.  Each scientific theory implies  ontological commitments or embodied presuppositions about what the world is like (Quine, 1980). 


How are we to deal with different ontological commitments associated with different knowledge claim made both by different fields of scholarship and from other people. The answer might simply be that each theory implies a set of “furniture” and in order to describe and classify things in the world you have to decide which theory is true (or most fruitful). As Quine claimed: There is “ no way to rise above the particular theories we have; no way to harmonize and unify their respective claims”.  To do classification or ontology is to be involved in the competition between different theories and views.


The furniture sets may be overlapping among different views.  This may be because the different views are not totally different – or have not yet been analyzed as clearly separated cases. Also the problem of incommensurability may be at play: When do we know that people are speaking of the same thing? They may understand each other’s words and signs differently. In spite of such problems theories may turn out to be fruitful, consensus may arise, and knowledge be stabilized (at least for a period), why epistemological problems may not hinder progress. 



4a. Disciplines and the social organization of knowledge

We have seen that “disciplines” and “phenomena” have competed about being the primary principle of division in classification.  Mills & Broughton (1977, p. 37) made a very clear argumentation for the use of disciplines as the basic organizing principle in the introduction to the Bliss 2 system. They wrote: 

           "5.55       Disciplines and phenomena     

           5.551         It should be clear from the last section (5.542) that although the disciplines reflect discrete systems of knowledge they yet share to some degree the same phenomena studied. The implication of this for a general classification is that the basic organization of information will subordinate material on a given phenomenon to the discipline or subdiscipline from whose viewpoint it is being regarded. So documents on the subject of the phenomenon "Color" for example, will not be kept together insofar as they will be assigned to the different disciplines (Physics, Art, etc.) their treatment reflects.                  

           5.552           However, it should be recognized that there is, theoretically, a quite different way of organizing a general classification. This would be to make the first division of the field of knowledge into phenomena (from subatomic particles to planetary bodies and stars, from single cells to particular organisms and particular societies, and so on) and to subordinate to each phenomenon the disciplinary aspects from which it may be treated; e.g. Color—in Optics, in biology, in Art, etc.; or, Food—in Agriculture, in Nutrition, in Cookery, in Economic resour­ces, etc.; or, Water—in Chemistry, in Geology, in Biology, in Engineering, in Transport, etc.   

           5.553           Such an arrangement would run counter to the way we usually study things and the way most information is marketed, which reflects the division of labor by discipline. There are relative few persons, if any, specializing in a given phenomena from all its aspects. Indeed, such a specialized study would require a training, which is at present hard to envisage. 

           5.554           Nevertheless, a growing number of documents do reflect a multi-disciplinary approach, although authorship of such works is usually, and not surprisingly, also multiple, as in the case of symposia. Such material poses a special problem for the older general classifications, which are sometimes called "aspect" classifications in that their basis of arrangement is by aspect or "discipline", not by phenomena. This does not, however, invalidate the general correctness of the decision they all reflect, which is to treat classification by discipline as being on the whole more helpful to users. It may be noted that the factual literature for children has always shown a strong tendency to con­centrate on phenomena rather than discipline—e.g., "the big book of trains" which considers most aspects of the railway system". (Mills & Broughton, 1977, p. 37).


The term “discipline” refers both to organizational units in educational programs (for example, in schools) and to organizational units in knowledge production. Journals, the primary communication channel in science, are mostly disciplinary units, and scientists are often supposed to compete within disciplinary borders.

"The scientific discipline as the primary unit of internal differentiation of science is an invention of nineteenth century society. There exists a long semantic prehistory of disciplina as a term for the ordering of knowledge for the purposes of instruction in schools and universities. But only the nineteenth century established real disciplinary communication systems. Since then the discipline has functioned as a unit of structure formation in the social system of science, in systems of higher education, as a subject domain for teaching and learning in schools, and finally as the designation of occupational and professional roles." (Stichweh, 2001, p. 13727).


Mattei Dogan puts forward the view that disciplines are no longer the most important units in scientific communication. He finds that no person in disciplines like sociology today can master the knowledge of the whole discipline and contribute to many of them. Each formal discipline gradually becomes unknown in its entirety. “The process of specialization has tended to disjoin activities which had previously been united, and to separate scholars belonging to the same formal discipline, but who are interested in different fields.” (Dogan, 2001, p. 14851).  Instead, the specialty is now the main source of academic recognition and professional legitimating.

In LIS have disciplines mostly been studied in bibliometrics and in scientific communication.  In bibliometrics have, for example, co-citation maps demonstrating a kind of intellectual structure in specialties, in disciplines and among disciplines been produced (see, e.g., White & McCain, 1998).  In scientific communication has models of the communication systems in disciplines been proposed (see, e.g., Fjordback Søndergaard; Andersen & Hjørland, 2003).


The core issue here is, however, that the study of disciplines has not in LIS been connected with the classification of disciplines as this is done in most systems, including the DDC. When library classification is primary understood as the classification of disciplines (and only secondarily with the classification of the phenomena studied by the disciplines) we should expect LIS to do research about disciplines.  Consider the following quote:


"Thus, between 1850 and 1945, a series of disciplines came to be defined as constituting an area of knowledge to which the name “social science” was accorded. This was done by establishing in the principal universities first chairs, then departments offering courses leading to degrees in the discipline. The institutionalization of training was accompanied by the institutionalization of research: the creation of journals specialized in each of the disciplines; the construction of associations of scholars along disciplinary lines (first national, then international): the creation of library collections catalogued by disciplines”. (Wallerstein, 1996, p. 30).


This quote claims that library classifications followed a classification of research and teaching programs at internationally leading universities.  This seems to imply that the updating of library classifications based on disciplines ought to be informed by some kind of empirical study of developments in the disciplinary structures in higher education. In a field which calls itself “science” (library and information science) it seems problematic that this has not so far been the case. ((A reason may be that libraries of practical reasons do not want to update their classifications. But then the classifications become obsolete and I addition LIS looses its ambition to be a research based field).

The kind of studies that can form a basis for the updating of LIS-classifications in relation to disciplines is related to studies such as Wallerstein (1996) although the first study seems to lack parts of the empirical and scholarly basis for its conclusions. The point here is 1) that there is very little research of this kind 2) it is very scattered and often related to the history and sociology of science and universities 3) that this scattered research is useful for knowledge organization why we in the field should know about it, should integrate it in our own research and contribute to expand it.  

5.    Epistemological paradigms in KO
1) The first edition of the Dewey Decimal Classification (DDC) came in 1876. Today this system is the most widely used library classification system internationally. In a way this system represents together with the UDC and similar systems the dominant approach to knowledge organization in libraries. See traditional approaches to KO for a further discussion of what kind of theory and method is involved in this approach.
2) Probably the most distinct approach to KO within Library and Information Science (LIS) is the facet-analytic paradigm (or analytico-synthetic approach) developed by S. R. Ranganathan (1892-1972) and further by the British Classification Research Group (1952-) Colon Classification and Bliss 2 are among the most important systems developed on this theoretical basis. This approach has still a strong position in the field, not least in London, but has relatively lost influence during the last decades. It is the most explicit and "pure" theoretical approach to KO, but it is not by implication necessarily the most important one. Principles from this tradition have increasingly influenced the development of classification systems, also old systems such as the DDC (cf., Miksa, 1998). The strength in this approach is it logical principles and the way it provide structures in KOS (classifications as well as thesauri, for example).
3) The 1950's saw the introduction of the computer and the concept of information storage and retrieval. This gave rise to computed based techniques (as well as semi-automatic techniques) for KO. The most primitive systems were KWIC-indexing-systems, the most influential systems were based on the vector space model developed by Gerald Salton. The information retrieval tradition (IR), which was founded in the 1950'ties with experimental traditions like Cranfield (later continued in the TREC-experiments and with the development of Internet search engines). The Cranfield experiments found that classification systems like UDC and facet-analytic systems were less efficient that free-text searches or low level indexing systems ("UNITERM"). Although KOS such as thesauri and descriptors are children of the IR-tradition, the main tendency has been to question the value of traditional classification and facet analysis and human indexing all together. It has more or less implicit worked with the assumption that algorithms working on textual representations (best full text representations) may fully substitute human indexing as well as algorithms constructed on the basis of human interpretations. If one does not question the results obtained in this approach it implies the end of knowledge organization as a research field to be substituted by IR. This is the reason why it is important to consider IR as one among other approaches to KO in order to identify its relative strengths and weaknesses. (See Information retrieval approach to KO).
    Besides, the development of online information systems gave rise to free-text searching and descriptor based indexing and other techniques which seriously challenged traditional forms of KO such as the UDC and facet analysis. These techniques are less distinct as an approach to KO compared, for example, to the facet-analytical paradigm. Among other things they blur the borders between KO and information retrieval (IR) as well as the borders between Library and Information Science (LIS) and fields like computer science and computational linguistics.
4) In 1963 Eugene Garfield introduced the Science Citation Index which gave rise to bibliometric knowledge organization. (See, for example, maps in Åström (2002) as examples of bibliometric maps: Bibliometric_MAP_LIS.PDF; Bibliometric_LIS_2.PDF).
5) In the 1970ties and 1980ties forms of user-based and cognitive approaches to KO developed. One of the prominent examples of systems developed from this approach is the Book House System developed by Annelise Mark Pejtersen and associates.
6) The 1990ties were influenced by new technologies such as the Internet and full-text databases. Today the semantic web is one of the dominant front-technologies in KO. These technologies may be viewed as a continuation of the IR-tradition (3).
7) The 1990ties saw also an increasing interest in social and interpretative approaches to KO including the formulation of domain analysis as an approach to LIS in general and to KO in particular. Domain analysis is an approach based on an explicit theory of knowledge.
8) Many other approaches have been suggested. Among them semiotic approaches, "critical-hermeneutical" approaches, discourse-analytic approaches and genre-based approaches. They are not going to be discussed further at this place, but the mentioned approaches can be seen as belonging to the same family to which also the domain-analytic approach belongs.

Relations to epistemology:
The facet analytic approach (with the method of logical division) is understood as a rationalist approach to KO.

Techniques based on "similarity" (e.g. statistical similarity measures) are understood as empiricist approaches to KO.

Organization based on historical developments and on relating concepts to the discourses in which they have been constructed is understood as a historicist approach to KO.

Organization of information based on explicated goals, values,  interests or consequences are understood as pragmatic approaches.