Agent-oriented Constructivist Knowledge Management

Abstract

In Ancient Times, when written language was introduced, books and manuscripts were often considered sacred. During these times, only a few persons were able to read and interpret them, while most people were limited in accepting these interpretations. Then, along with the industrial revolution of the XVIII and XIX centuries and especially boosted by the development of the press, knowledge slowly became available to all people. Simultaneously, people were starting to apply machines in the development of their work, usually characterized by repetitive processes, and especially focused in the production of consuming goods, such as furniture, clocks, clothes and so on. Following the needs of this new society, it was finally through science that new processes emerged to enable the transmission of knowledge from books and instructors to learners. Still today, people gain knowledge based on these processes, created to fulfill the needs of a society in its early stages of industrialization, thus not being compatible with the needs of the information society. In the information society, people must deal with an overloading amount of information, by the means of the media, books, besides different telecommunication and information systems technology. Furthermore, people’s relation to work has been influenced by profound changes, for instance, knowledge itself is now regarded as a valuable work product and, thus, the workplace has become an environment of knowledge creation and learning. Modifications in the world economical, political and social scenarios led to the conclusion that knowledge is the differential that can lead to innovation and, consequently, save organizations, societies, and even countries from failing in achieving their main goals. Focusing on these matters is the Knowledge Management (KM) research area, which deals with the creation, integration and use of knowledge, aiming at improving the performance of individuals and organizations. Advances in this field are mainly motivated by the assumption that organizations should focus on knowledge assets (generally maintained by the members of an organization) to remain competitive in the information society’s market. This thesis argues that KM initiatives should be targeted based on a constructivist perspective. In general, a constructivist view on KM focuses on how knowledge emerges, giving great importance to the knowledge holders and their natural practices. With the paragraph above, the reader may already have an intuition of how this work faces and targets Knowledge Management, however, let us be more precise. Research in Knowledge Management has evolved substantially in the past 30 years, coming from a centralized view of KM processes to a distributed view, grounded in organizational and cognitive sciences studies that point out the social, distributed, and subjective nature of knowledge. The first Knowledge Management Systems (KMSs) were centrally based and followed a top-down design approach. The organization managers, supported by knowledge engineers, collected and structured the contents of an organizational memory as a finished product at design time (before the organizational memory was deployed) and then disseminated the product, expecting employees to use it and update it. However, employees often claimed that the knowledge stored in the repository was detached from their real working practices. This led to the development of evolutionary methods, which prescribe that the basic KM system is initially developed and evolves proactively in an on-going fashion. However, most of the initiatives are still based on building central repositories and portals, which assume standardized vocabularies, languages, and classification schemes. Consequently, employees’ lack of trust and motivation often lead to dissatisfaction. In other words, workers resist on sharing knowledge, since they do not know who is going to access it and what is going to be done with it. Moreover, the importance attributed to knowledge may give an impression that these central systems take away a valuable asset from his or her owner, without giving appreciable benefits in return. The problems highlighted in the previous paragraph may be attenuated or even solved if a top-down/bottom-up strategy is applied when proposing a KM solution. This means that the solution should be sought with aim at organizational goals (top-down) but at the same time, more attention should be given to the knowledge holders and on the natural processes they already use to share knowledge (bottom-up). Being active agency such an important principle of Constructivism, this work recognizes that the Agent Paradigm (first defined by Artificial Intelligence and more recently adopted by Software Engineering) is the best approach to target Knowledge Management, taking a technological and social perspective. Capable of modeling and supporting social environments, agents is here recognized as a suitable solution for Knowledge Management especially by providing a suitable metaphor used for modeling KM domains (i.e. representing humans and organizations) and systems. Applying agents as metaphors on KM is mainly motivated by the definition of agents as cognitive beings having characteristics that resemble human cognition, such as autonomy, reactivity, goals, beliefs, desires, and social-ability. Using agents as human abstractions is motivated by the fact that, for specific problems, such as software engineering and knowledge management process modeling, agents may aid the analyst to abstract away from some of the problems related to human complexity, and focus on the important issues that impact the specific goals, beliefs and tasks of agents of the domain. This often leads to a clear understanding of the current situation, which is essential for the proposal of an appropriate solution. The current situation may be understood by modeling at the same time the overall goals of the organization, and the needs and wants of knowledge holders. Towards facilitating the analysis of KM scenarios and the development of adequate solutions, this work proposes ARKnowD (Agent-oriented Recipe for Knowledge Management Systems Development). Systems here have a broad definition, comprehending both technology-based systems (e.g. information system, groupware, repositories) and/or human systems, i.e. human processes supporting KM using non-computational artifacts (e.g. brain stormings, creativity workshops). The basic philosophical assumptions behind ARKnowD are: a) the interactions between human and system should be understood according to the constructivist principle of self-construction, claiming that humans and communities are self-organizing entities that constantly construct their identities and evolve throughout endless interaction cycles. As a result of such interactions, humans shape systems and, at the same time, systems constrain the ways humans act and change; b) KM enabling systems should be built in a bottom-up approach, aiming at the organizational goals, but understanding that in order to fulfill these goals, some personal needs and wants of the knowledge holders (i.e. the organizational members) need to be targeted; and c) there is no “silver bullet��? when pursuing a KM tailoring methodology and the best approach is combining existing agent-oriented approaches according to the given domain or situation. This work shows how the principles above may be achieved by the integration of two existing work on agent-oriented software engineering, which are combined to guide KM analysts and system developers when conceiving KM solutions. Innovation in our work is achieved by supporting topdown/bottom-up approaches to KM as mentioned above. The proposed methodology does that by strongly emphasizing the earlier phases of software development, the so-called requirement analysis activity. In this way, we consider all stakeholders (organizations and humans) as agents in our analysis model, and start by understanding their relations before actually thinking of developing a system. Perhaps the problem may be more effectively solved by proposing changes in the business processes, rather than by making use of new technology. And besides, in addition to humans and organizations, existing systems are also included in the model from start, helping the analyst and designer to understand which functionalities are delegated to these so-called artificial agents. In addition to that, benefits as a result of the application of ARKnowD may be also attributed to our choice of using the proper agent cognitive characteristics in the different phases of the development cycle. With the main purpose of exemplifying the use of the proposed methodology, this work presents a socially-aware recommender agent named KARe (Knowledgeable Agent for Recommendations). Recommender Systems may be defined by those that support users in selecting items of their need from a big set of items, helping users to overcome the overwhelming feeling when facing a vast information source, such as the web, an organizational repository or the like. Besides serving as a case for our methodology, this work also aims at exploring the suitability of the KARe system to support KM processes. Our choice for supporting knowledge sharing through questioning and answering processes is again supported by Constructivism proponents, who understand that social interaction is vital for active knowledge building. This assumption is also defended by some KM theories, claiming that knowledge is created through cycles of transformation between two types of knowledge: tacit and explicit knowledge. Up to now, research on KM has paid much attention to the formalization and exchange of explicit knowledge, in the form of documents or other physical artifacts, often annotated with metadata, and classified by taxonomies or ontologies. Investigations surrounding tacit knowledge have been so far scarce, perha

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 44 citations exist upstream.

Papers that cite the same references — the relational fallback while embeddings are still backfilling.