Words Theoretical development of information science: a brief history Birger Hjørland
“The systems oriented paradigm” and “the subject knowledge view”
Download 225.5 Kb.
|
- Navigate this page:
- 5. Cognitive theory
4. “The systems oriented paradigm” and “the subject knowledge view” We have seen that there were great expectations to Shannon’s theory, and Wersig wrote: One could call the developmental stage [of information science] from1948 to the 1970s the 'Shannon and Weaver phase', because most of the discussions and attempts to structure the concept of information relied on the reception of Shannon via Weaver (Wersig, 2003, p. 313). Following the 'Shannon and Weaver phase' Wersig presents “the cognitive viewpoint”. But before we turn to that, it should be said that an important issue seems to be lacking in his outline (which may be due to a neglect of considering development in practice). The period from the beginning of the 1970s to about 1990 was a period in which online information services developed rapidly and influenced information science and information professionals very much. Such online services revolutionized library reference services and from my point of view such services is very much what information science and information profession is about. As explained in Hjørland (in press, a) information specialists developed roles as expert searchers and contributed to the design of online databases (including knowledge of thesauri, natural language, citation searching, etc.). This development cannot be considered an application of Shannon’s theory and must therefore be based on some alternative framework (to be discussed below). At the research front around 1970 – and important to this day – was a paradigm developed by the so-called Cranfield-experiments (see Cleverdon, Mills and Keen, 1966; Cleverdon, 1970), which was later labeled “the physical paradigm” (e.g. by Ellis, 1992a, b) or “the systems centered approach” (e.g. by Saracevic, 1999). The experimental environment was Cranfield College of Aeronautics and the tests were carried out in the same way as other technological products are tested (hence the name “physical paradigm”). The physical paradigm is derived from the importation of the scientific method into LIS. It views information as flowing from a source to a destination, the user, who is relatively passive. The focus is on the process of translation information into a message that the system can convey to the user, as suggested by Claude Shannon’s Communication Theory. …
The Cranfield studies introduced the famous measures “recall” and “precision” in order to evaluate the efficiency of different systems and search strategies – which gave rise to an enormous literature about the concept “relevance”. The simple idea is that all relevant documents should be retrieved while non-relevant documents (noise) should be avoided. All three terms are still core terms in information science, but it is an issue whether they should be based on scientific criteria (and expert evaluations) or on user-based criteria and on user-evaluations1. In the Cranfield experiments, the evaluation of systems and search strategies were done by subject specialists2, whereas the later user-based approach let the users make the evaluation3. The most explicit scientific criteria of relevance and recall have been developed in evidence based medicine, cf. Hjørland (in press a). Olson and Boll’s quote above said that the Cranfield experiments used “relatively arbitrary notion of what a successful search might be”. I do not agree in this statement (and not either in view that the physical paradigm should be understood as based on Shannon’s theory). In a former publication I described how evaluations were actually carried out: Cleverdon (1970) reanalyzed some results from the Cranfield II experiments. The types of search questions discussed were both “realistic” or “real-life questions” and “prepared questions” (which is surprising, given the description of this view from the user-oriented community). Relevance assessments were made by people with different backgrounds, mostly scientists in the field. Each assessor evaluated each document (in full text) on a five-point scale and made qualitative notes about the assessment. Most important is that relevance was evaluated in relation to its possible function for the user because this is directly opposed to how the systems view is mostly being described. The paper further discussed how relevance assessments vary greatly among different assessors. Appendix 1 in Cleverdon (1970) lists the test-questions and the real documents used in the test. This seems important because it makes interpretations of the relevance-assessments possible. This procedure seems different from how it is described by the user-oriented researchers (Hjørland, 2010, p. 220). I do not see this procedure as a “relatively arbitrary notion of what a successful search might be” as quoted above, but it is correct that the view of that time was mostly based on a positivist assumptions. What was not investigated was how different “paradigms” in the field have different evaluation criteria (and a discussion of on which subject knowledge view evaluations should be performed). The reservations raised by Olson and Boll in relation to the physical paradigm may be seen as a justification to their next section, which is: “The cognitive paradigm”. However, if the cognitive paradigm does not in reality make better solutions for users – as I claim I fail to do – then Olson and Boll’s reservations posed against the physical paradigm fall flat. Information science started with the underlying premise that information specialists needs subject knowledge (just as we saw that Shera, 1968, argued for subject knowledge in librarianship4). The term “information scientist” was first used by Farradane (1953) two years before the term “information science” was first used and was used specifically to denote a scientist trained to help other scientists in finding information: the qualifications were a degree in science or engineering, a second language and 5 years’ experience in information work. The Institute of Information Scientists established 1958 reflected this definition (see Vickery and Vickery, 1987, pp. 361-369). Robinson and Bawden (2013) also underlined the subject-based nature of early British information science. It seems therefore that information science in practice was founded on “the subject knowledge view”, somewhat related to what was later named “the domain-analytic view” by Hjørland and Albrechtsen (1995). The important difference may be that the subject knowledge view that time was mainly “positivist” (assuming consensus among researchers) while the theory of knowledge today is more influenced by paradigm theory (assuming conflicting values and dissensus in “information”). In other words: The subjectivity in relevance assessments in both experts and users should be examined, and perhaps the best explanation in the variations is to be found in the different theories and paradigms to which the assessors subscribe. “The systems centered approach” is, by the way, a bad term because it not just about computers, but also about the documents represented in the databases, their different genres, terminology, relevance, properties, subject access points etc. (and in this respect can be seen as a continuation of documentation as described above). I do not see information science as primarily about human-computer interaction but about human interaction with mankind’s recorded knowledge and culture (via computers). It can be mentioned, for example, that Science Citation Index was among the important bibliographic databases developed (in 1963). This gave rise to the study of how researchers and documents are connected in citation networks. It also provided basis for coupling information science with the sociology and philosophy of science (e.g. studying how disciplines differ, motivations to cite other papers etc.). This field (bibliometrics) has today developed into one of the most important research fields in information science (as well as other fields), and because its links with sociology and philosophy, it has potentials to be an important part of the theoretical foundation for information science (although that field too is often based on positivist assumptions). From this perspective information science is a metascience concerned with the optimization of infrastructures and scholarly communication in different domains (cf., Talja, 2010; Hjørland, in Press, b). 5. Cognitive theory ‘Cognitive science’ or ‘cognitive sciences’ is an interdisciplinary research field with roots in cognitive psychology, linguistics, artificial intelligence and philosophy. It is an offshoot of Shannon’s information theory along with fields like cybernetics and control theory, and the development of computer technology. The field was established around 1975 with Norman and Rummelhart’s Explorations in Cognition (1975) being among the first books about it. The journal Cognitive Science started in 1977 and The Cognitive Science Society held its first annual meeting in 1979. Cognitive theory, or more precisely, the computational theory of mind, as it developed in this movement represents a particular kind of functionalism in which cognitive processes (in humans, animals and computers) may be described as information processing. All behavior is understood as rule-based (implemented in programs). Units of information take the form of mental representations. The computational theory of mind redefined human beings as processors of information, whereby they opened the way for researchers to rigorously study mental processes using the tools of science and computer technology, enabling them to make thinking visible and open to experiments. It is therefore also characteristic of cognitive science that theories and models about human cognition may be tested on computers and vice versa, and that new developments in computer science inspired new models of human cognition. Much of the technical language used in cognitive theory is appropriated from information theory and computer technology. An important part of the theory is the dichotomy between hardware and software – and that software may run on different kinds of hardware (brains or computers). One of the underlying assumptions is also that the study of general psychology can be used to develop computers which imitate human intelligence (‘artificial intelligence’). Like Shannon’s information theory, cognitive science was received with extraordinary optimism and excitement in the beginning. Philosophically it represented a rationalist turn: The philosophical antecedents of cognitive theory are numerous, but one philosopher stands out as having a profound effect on the assumptions made by many cognitive theorists. René Descartes (1596–1650) is often referred to as a rationalist philosopher, meaning that he focused upon the workings of the mind. … Descartes posited the notion that humans have innate ideas and that there is a duality between mind and body. These two ideas were to be seminal for cognitive theory. They, along with the computer metaphor, significantly directed the course cognitive theory would take and would be the source of many of the critical debates concerning it (Faux, 2014, p. 250).
There has been much criticism of the view that humans are information processing devices. The criticism seems different from the criticism of Shannon’s theory where everybody seemed to agree that this would be a fruitful theory if it was just used in its proper domain. Although this may also be the case with the computational theory of mind, the debate here seems to be much more existential: whether or not it is a fruitful theory of human cognition at all. We saw above that there is a link between the computational theory of mind and rationalism. The classical opposition to rationalism came from empiricism. Empiricism does not assume innate ideas, but is based on the view that humans are born with a tabula rasa, an empty mind. Empiricism is related to associationism and behaviorism, and based on these alternative philosophical perspectives the computational approach known as neural networks was developed. The theory (and technology) of neural networks represents an alternative theory that does not assume the existence of mental representations, as cognitive theory does. However, we shall not go deeper into behaviorism and neural networks because both rationalism and empiricism are based on an individualist philosophy, which in my view represents a trap which can be avoided if we take a historicist and pragmatic point of departure. Before we turn to that, it should be emphasized that the point raised in this paragraph is extraordinary: that core issues related to the development of today’s front-line technologies are based on classical philosophical issues. Cultural psychologist Carl Ratner wrote about ‘the psychological fallacy’ which is as relevant to cognitivism as it is to other alternatives to cultural psychology: In 1910 Dewey wrote a statement that expresses a central tenet of cultural psychology. He said that the processes that animate and form consciousness lie outside it in social life. Therefore, the objective for psychologists is to use mental phenomena (e.g., perception, emotions) as clues for comprehending the life processes that they represent. […] ‘The supposition that these states [of consciousness] are somehow existent by themselves and in this existence provide the psychologist with ready-made material is just the supreme case of the ‘psychological fallacy’’ (Ratner, 2002, p. 3). Two of the major critics of the computational theory of mind were also two of cognitive theory’s architects. Ulrich Neisser (1976) criticized cognitive theory for being too focused upon laboratory research, arguing that it lacks face validity because of this; thus, its applicability to the study of how people think and solve problems in real-world settings is limited at best. To overcome this, research needs to be conducted in real-world settings. Jerome Bruner (1990) has criticized cognitive theory for becoming technicalized (p. 4): cognitive theory has fixated on how people process information at the expense of understanding how we construct meaning. Bruner thus shifted the perspective from cognitivism to cultural psychology. ‘Through a cultural lens, the notion of abstract, individual processors of information becomes a shallow and one-dimensional view of the very deep and complex set of activities we call thinking’ (Faux, 2014, p. 251). Faux also summarizes the critique of cognitive theory by social constructionists, postmodernists and humanists. These critiques converge on the notion that we are embodied beings in the world and that our thinking is a function, not of information processing, but of social interaction in cultural contexts. In an article in Human IT the Swedish philosopher Peter Gärdenfors noted that
Gärdenfors thus put forward a view related to that of Frohmann (2004) that the study of information has to start with human activities, and that the approach used in cognitive science and in information science has to be turned upside down. This is also the view that informs the present author. 5.2. The cognitive view in information science The above section has focused on the cognitive view in general. We shall now focus on the discourse about the cognitive view in information science. The cognitive viewpoint in IS was initially formulated by Brookes (1980)5, Belkin and colleagues (Belkin, 1984, 1990; Belkin, Oddy and Brooks, 1982) and Ingwersen (1982, 1992). Core theoretical assumptions seem to be:
Two concrete examples of systems claimed to be based on the cognitive approach are the Book House (developed in information science) and WordNet (developed in cognitive science, not within information science). The analysis of such concrete achievements seems important in order to understand a given approach. Hjørland (2013e) made an examination of these two systems and found that it is unclear which role the study of users has in reality played in their development. In spite of their explicit basis in cognitivism, they may actually largely be based in subject knowledge. Sanna Talja, Kimmo Tuominen and Reijo Savolainen provided an examination of different views in information science in which they placed ‘the cognitive view’ under ‘constructivism’ (more specifically under ‘cognitive constructivism’) and wrote: In IS [information science], constructivist ideas are commonly labelled under ‘the cognitive viewpoint’. The cognitive viewpoint in IS […], does not represent cognitivism, however. Cognitivism is an approach that significantly informed artificial intelligence in drawing straightforward analogies between human information processing and computing (Ingwersen, 1992, pp. 19–25, 227). The cognitive viewpoint in IS differs from cognitivism by laying major emphasis on the way in which knowledge is actively built up by the cognising subject, that is, by the individual mind to serve the organisation of internal and external reality (Talja, Tuominen and Savolainen, 2005, p. 81). I agree with Talja, Tuominen and Savolainen that it is unclear what the cognitive view stands for because it seems not to be a consistent position. Before 1991 the cognitive view did in my opinion represent cognitivism and the attempt to draw analogies between human information processing and computing and to provide cognitive models for information transfer such as Belkin’s (1984) MONSTRAT, Ingwersen’s MEDIATOR6 Later on Ingwersen (1992) made a distinction between cognitivism and his own view, labelled ‘the cognitive view’ (probably due to criticism raised by Hjørland, 1991). And later on again Ingwersen (1996) suggested a “holistic cognitive framework for IR”. In his 1992 monograph, Ingwersen (p. 157) saw the cognitive view as a synthesis between user-oriented approaches and the “traditional approach” and wrote: “The transformation from the user-oriented and the traditional approaches into a cognitive one happens when IR research comes to have each other’s isolated models in mind.” In Ingwersen and Järvelin (2005, 191), however, user-oriented and cognitive views seem no longer to be separated. Here, the authors “[discuss] the development of cognitive and user-oriented research from the 1970s and onwards under one umbrella” and state that “the cognitive approach to IR could briefly be characterized as user- and intermediary-oriented.” I interpret this—in line with other writings—as a tendency to give up the cognitive approach as differentiated from user-based approaches. The cognitive view in 2005 now claim to have turned from an individualist to a social perspective: The present book … reflects a further development of the cognitive viewpoint … by providing a contextual holistic perspective. The quite individualistic perspective laid down in the former monograph [Ingwersen, 1992] is hence expanded into a social stance including generation, searching and use of information” (Ingwersen and Järvelin, 2005, p. vii). Although the cognitive view started as a reaction to the systems-oriented view (including IR research and bibliometrics), the latter seems to be the approach that Ingwersen has mostly later turned towards. “The cognitive view” in Ingwersen’s research seems therefore to have missed its characteristic as a distinguishable approach, while other researchers seems to continue the original cognitivist profile (e.g., Jörgensen, 2003; Todd, 2005; Zhang, Liu, and Belkin, in press). Therefore, the cognitive view may not really be considered a unified theory or tradition. Of this reason the question of whether or not cognitivism in information science is related to the constructivist position of Jean Piaget and others, as suggested by Talja, Tuominen and Savolainen (2005) cannot be answered with yes or no, but only be answered by analyzing specific suggestions that have been put forward. Belkin’s ASK hypothesis (Belkin, 2005), for example, should it be considered constructivist rather than cognitive? We shall not answer that question here, because both views share an individualist approach, which I find problematic. It is also outside the scope of this paper to provide introductions to and discussions of theoretical positions such as constructivism, constructionism, discourse analysis etc. Many of the assumptions of the cognitive view turn out to be problematic; rather than openly recognize or debate such problems, the strategy of some cognitivist researchers seems to be to try to make the label fit whatever approach seemed to be proposed in the community of information science. Tefko Saracevic wrote about two traditions of research in information retrieval: a systems-centered tradition (in computer science) and a user-centered tradition (in information science): The split is not only conceptual, looking very differently at the same process, but also organizational. The systems centered side is now mostly concentrated in the Special Interest Group on Information Retrieval (SIGIR) of the Association for Computing Machinery (ACM), while the user-centered cluster congregates around the American Society for Information Science (ASIS). Each has its own communication outlets—journals, proceedings, and conferences. There is less and less overlap of authors and works between the two outlets. We have two camps, two islands, with, unfortunately, relatively little traffic in-between. (Saracevic, 1999, p. 1057). In this quote Saracevic assumes that the research on IR taking place in information science (the user-centered tradition) has something important to offer. However, a much more pessimistic interpretation was put forward by Wersig: Information science has not reached a stage of development where it relies on a sufficiently sound theoretical and methodological base to be accepted at least as an important field of study. It is still looking for its identity (Vakkari and Cronin, 1992) and one of the reasons is that its representatives have never been able to demonstrate that they are pursuing something really different from computer science, due to the strong retrieval component, which remains one of the main features of recent information science (Wersig, 2003, p. 314). Today we all know that IT research in computer science has produced important tools such as Internet search engines. However, in the period dominated by the cognitive view, the field ‘has not reached a stage of development where it relies on a sufficiently sound theoretical and methodological base to be accepted at least as an important field of study’ (Wersig, 2003, p. 314). If this is true (as I believe it is), it is indeed a serious situation that demands a reconsideration of the theoretical basis of the field. Sanna Talja’s view in the following quotation corresponds to my own evaluation of cognitivism and the cognitive view in information science: It is widely recognized that both individual information needs and institutional information access are socially conditioned. However, conducting information seeking research on a macro-sociological level has turned out to be difficult within the cognitive viewpoint, since it is basically a theory of how individuals process information. The cognitive viewpoint offers no concrete and obvious solutions to the question of how to conceptualize and study the socio-cultural context of information processes (Talja, 1997). The criticism of the cognitive view in information science is thus strong, and its influence seems to be declining7. What replaces it? Download 225.5 Kb. Share with your friends:
|