The study examines different metatheories of information and knowledge formation. By "metatheories" I refer to the often unspoken premises or abstract thinking models upon which empirical research and theorisation in Information Science (IS) is based. Metatheories are orientation strategies to the world. They give researchers a way to speak and think about phenomena under study. A metatheory makes it possible for information scientists to observe, for example, people's information behaviour or information systems from a certain angle or perspective. Thus, disparate metatheories allow different, even contrasting, versions of the research objects to be built. Each metatheory opens the world of understanding in one way but leaves it closed in others. Therefore, information scientists should reflect more on their metatheories.

My aim is not just to identify the most used metatheories in IS but also to discuss their applicability in building digital libraries (DLs). By drawing on the Sense-Making approach developed by Brenda Dervin and the emerging research tradition of Social Informatics I assume that our metatheories affect the way we develop and design DLs. Thus, metatheories are not just abstract thinking models. They also have practical consequences in how we choose materials for DLs and how we build their information architectures. Therefore, it is important to analyse the implications of different metatheories on the ways in which documents and other materials are arranged (classified or indexed) in DLs.

Two prevalent metatheories in Information Science (IS) are identified in the study: the information transfer model and constructivism. My aim is also to build and present an alternative to these orientation strategies. The metatheory directing my critique of both information transfer model and constructivism is (social) constructionism.

The information transfer model has been the most commonly used metatheory in IS. In this model information is seen as an abstract entity that is possessed by the sender of the message. The sender is an expert who forms knowledge by observing the world. The model assumes that knowledge, constructed by observing a phenomenon, reflects it like a mirror. The task of the sender is to transport this mirror intact from his mind to the mind of the receiver. Because direct thought transference is not possible, one needs a mechanism to carry the message. Thus, language functions in the model like a conduit in which the "information brick" can be put in order to be transported between the sender and his audience.

I criticise this model of an authoritarian worldview that supports expert power in naming and defining the world to others. When information is presupposed to be apolitical, neutral and objective one cannot see, for example, how language is used rhetorically and functionally in explicit or implicit conflicts between social actors or different interest groups.

Constructivism rests on the same subject—object dualism as does the information transfer model. However, it does not conceive of information as an objective but as a subjective "material". Constructivists assume that there is no information or knowledge outside the human mind or brain. The message being transported between minds A and B is just potential information. When information is absorbed by the mind of the receiver it transforms into real information or knowledge. There is no way in which the sender can ensure that the subjective sense-making process taking place in the mind of the receiver is similar to his own. Two different constructivist approaches are dealt with in the study, namely, the cognitive viewpoint and subjectivism.

The main problem with constructivism lies in its way to see the relationship between discourse and cognition. The assumption about the measurability of the changes in mental models through discourse is deconstructed with ideas derived from C. Wright Mills, Ludwig Wittgenstein and discursive psychology. Although constructivism does not see information as objective, it assumes – as does the information transfer model – that potential information is apolitical and neutral. Thus, constructivism does not allow information scientists to see the struggles about the status and authority of different ways of knowing in a society. Because constructivism tries to penetrate inside the recesses of individual minds it easily hides from view all the socio-discursive processes of knowledge formation.

Social constructionism does not assume information or knowledge to be either subjective or objective. Rather, it understands knowledge formation contextually and dialogically. Knowledge is a negotiated discursive construct that is created between people. Constructionists are interested in the rhetorical methods by which knowledge is created and supported in different conversations and conversational traditions. Symmetrically, constructionist analyses also deal with the discursive means that are use to deconstruct the factuality of versions about the phenomena under discussion. Constructionism overcomes the authoritative worldview of the information transfer model as well as the mentalistic assumptions of constructivism. This metatheory allows us to see how thoroughly rhetorical and interested factual talk and writing often actually is.

To explore the issue of constructionism and its implications further, constructionist discourse analysis was undertaken in the study to examine in detail the interview data collected from twenty patients waiting for heart surgery and their spouses. It is shown how different interpretative repertoires, i.e., ways of knowing or discourses, are constructed and deconstructed in the conversations between the interviewer and the interviewees. Chronically ill people can use many kinds of evidence to support their claims or versions. For example, they can rest on the discourse based on a lay bodily experience when contradicting or deconstructing medical views. They can also stress that as ordinary persons they cannot know what is wrong with them. In this context they often also emphasise that by penetrating the surface of their symptoms with technical instruments physicians can make exactly correct diagnoses.

I explain this kind of variation in talk by analysing the contextual functions of the different, even contrasting, versions being built in the interview situation. It is necessary to deconstruct medical knowledge when arguing against the moral ideology of health preservation that draws its authority from medical science. In this context, the speakers' own bodily experience weighs more than the cognitive authority of medical specialists. However, when talking about the diagnosis of one's own heart disease one needs to take into account the morally delicate issue of symptom recognition and diagnosis. The interviewees are sensitive to the potential claim that they have consulted the physician too late about their symptoms. Thus, it is necessary for the interviewees to stress that as lay persons they could not be exactly sure what was wrong with them before the diagnosis was made.

The results of the empirical analysis are used in considering how the interpretative repertoires identified can be articulated and expressed in DLs that are built for chronically ill people and that draw on different metatheories. I also examine what kind of information architectural solutions the metatheories analysed are connected with.

The DL based on the information transfer model concentrates on medical knowledge as well as on morally weighted suggestions concerning the dietary regimens and the general way of living of heart patients. In this context, the information architecture of the DL is most probably based on standard classification systems and controlled vocabularies. These systems and vocabularies are indispensable in digital as well as in physical libraries. However, in such DLs the information architecture is based on nouns, not on solutions that combine nouns and verbs.

The basic information architecture of constructivist DL would be quite similar with that based on the information transfer model. Yet, at the level of the user interface there would be a significant change. The constructivist DL would give each user an opportunity to personalise the user interface to suit his subjective needs.

In the above-mentioned DLs, the discourses about bodily experiences or anxieties of the users would fell into a subordinate position when compared to the views of the medical experts. In a constructionist DL this would not be the case. This kind of DL would not be an information system but a participatory system in which users would function as content providers and participants in conversations. Thus, users would not act outside but inside the DL. They could create documents for the DL as well as link relevant material to the collection. This kind of DL would celebrate the multiplicity of perspectives on the subject matters it deals with. It would not give a primary position to any particular perspective, voice or discourse.

The major innovation that constructionism brings to the information architecture of the DL is that the providers of the service would not just try to describe the information content of the resources the DL contains. The noun-based concepts normally used to depict documents would be integrated with verbs that would show the conversational relationship or context of documents and concepts. These verbs would represent different argumentative or non-argumentative moves in the conversational knowledge-creating processes. They would not just help the users to search for information and documents but also to make better sense of the subject matter the DL deals with. In the constructionist DL users would be empowered to add metadata to the documents. By using this metadata, users would be able to show relationships between different perspectives, documents and conversational traditions.