Chapter 4

Creative Conversation

This chapter explores the creative conversation from which human collective intelligence is emerging in the new digital communication environment and looks at how it functions and possible improvements. Creative conversation is the fundamental engine of knowledge communities, that is, communities seen from the perspective of their cognitive functioning. The first main idea put forward in this chapter is the inseparability of collective intelligence and personal intelligence. This idea is expressed in practical terms in the dialectical interdependence of social and personal knowledge management. Second, I stress the growing role of creative conversation in explicating, accumulating and organizing knowledge in the shared memories of knowledge communities. The chapter concludes with a third key idea: that the technical and social conditions for the collaborative construction of memory on the Web force us to radically rethink our traditional ways of organizing archives. Memory beyond the Web calls for a new symbolic medium for creative conversation, an open, universal, democratic and computable semantic sphere.

4.1. Beyond “collective stupidity”

Since the publication of my book Collective Intelligence in 1997¹, I have continually met with the classic (and, in my opinion, weak) objection that it is individual humans who are intelligent, while groups, more or less organized communities and, even more so, crowds are for the most part stupid. What are we talking about here? The term collective intelligence can have many different meanings, but all these meanings involve the combination of two concepts: cognition (“intelligence”) and society or community (“collective”). Cognition here is, very classically, the activity of perceiving, remembering, problem solving, learning, etc. Collective intelligence therefore refers to the cognitive capacities of a society, community or collection of individuals. This collective cognition can be seen from the perspective of the two complementary aspects of the dialectic between individual and society. On the one hand, the individual inherits and benefits from the knowledge, institutions and tools accumulated by the society he or she belongs to. On the other hand, distributed processes of problem solving, decision-making and knowledge accumulation emerge from conversations and, more generally, symbolic interactions among individuals.

With regard to inherited intelligence, it should be noted that individual cognitive capacities are almost all based on the use of tools – symbolic (languages, writing systems, various social institutions) or material (instruments of measurement, observation and calculation; vehicles and transportation networks; etc.) – that individuals have not invented themselves but that have been transmitted or taught to them by the surrounding culture. I have emphasized this enough in the previous chapter. Most of the knowledge used by those who claim that intelligence is purely individual comes to them from others, through social institutions such as the family, school or media, and this knowledge could not have been accumulated and developed without long intergenerational chains of transmission.

With regard to emergent cognition, it should be noted that the most advanced contemporary societies are based on institutions whose main engine is precisely collective intelligence in the form of well-ordered conversation: these include democracy, the market and science.

The principles of democracy do not guarantee that inept or corrupt leaders will never be elected or that extremist or violent policies will never be adopted by the majority of a population. Universal suffrage, political pluralism, the balance of powers, freedom of expression for all and respect for human rights in general (and those of minorities in particular) are, however, more conducive to civil peace and human development than dictatorships or regimes dominated by a single party or a closed group of the privileged few. In democracy, collaborative intelligence comes about, not as a result of the majority imposing its will, but rather, out of the decisions of voters or the members of various parliaments after open deliberation during which different views can be expressed and responded to^2.

The existence of a free market regulated by law will not prevent economic crises or income inequalities. Historical experience, however, shows that planned economies in which a small number of bureaucrats decide the orientations of production and set prices are much less efficient than market economies, in which producers and consumers as a whole contribute – imperfectly and with all the attendant distortions – to deciding prices and levels of production and consumption³. Here, creative conversation is ideally an economic negotiation informed by realities and respectful of laws. I note, in order to avoid any misunderstanding, that this perspective is open to government interventions aimed at making markets more dynamic and more conducive to human development, such as through the construction of infrastructure, the creation of circumstances favorable to education and research, or the implementation of social assistance programs.

Finally, the scientific community is governed by principles of collective intelligence such as peer evaluation, reading and citing of colleagues, reproducibility of observations and sharing of data. None of these principles prevent repetitive mediocrity, errors or “false” theories. Conversation by the scientific community, conversation that is both collaborative and competitive, is obviously preferable, for the advancement of knowledge, to arguments from authority or hierarchical, dogmatic, opaque institutions with inquisitorial powers.

More recently, the success of the open software movement, which is based on the free collaboration of programmers worldwide, and the multilingual online encyclopedia Wikipedia, in which authors, readers and editors exchange roles to further the dissemination of knowledge, are striking examples of the power of collective intelligence emerging from a civilized creative conversation.

Thus the facile irony about collective stupidity (which is obviously always the stupidity of “others”) fails to recognize all that our individual wisdom owes to tradition and that our most powerful and useful institutions owe to our ability to think and decide together. Need I add that my emphasis on the collective aspect of human intelligence in no way implies the abdication of critical thought or individual originality? The concept of collective intelligence for which I am arguing here is the opposite of conformism or sterile standardization. The full recognition of what we owe to the traditions or communities we are part of implies precisely the moral obligation to enrich the common good through original, relevant creative effort. Collective intelligence can only be productive by combining or coordinating unique elements and facilitating dialog, and not by leveling differences or silencing dissenters. Finally – need it be repeated? – no common knowledge can be created, accumulated or transmitted without an individual effort to learn.

4.2. Reflexive explication and sharing of knowledge

4.2.1. Personal and social knowledge management

4.2.1.1. Introduction to knowledge management

Most of us no longer live, as our ancestors did, in a single tribe. Contemporary social life generally has us participate in many communities, each with a different cultural tradition or knowledge ecosystem. Members of a family, speakers of a language, citizens of a city or nation, followers of a religion, practitioners of a discipline, learners of a technique, amateurs or masters in an art, collaborators in a business or organization, fans of a TV show or video game⁴, members of a thousand networks, associations or working groups, we participate in more than one cultural community. If we look at these communities from a cognitive perspective, they are constituted through an autopoietic process of construction, reproduction and transformation of knowledge ecosystems. These are “working” communities in the information economy or, if you will, social learning enterprises. Their creative conversations accumulate, manage and filter memories in which collective identities and personal identities define each other, and the capacity for thoughtful interpretation and the capacity for informed action answer each other. For each of these communities, the maintenance and use of its knowledge capital, or the management of its knowledge, is thus a major concern.

Since I am going to use the now classic term knowledge management (KM), I would like to prevent any misunderstandings at the outset⁵. It is generally agreed that the only things that can be “managed” objectively and rationally are data, in particular digital data. On the other hand, it is still possible – but rather more difficult – to manage the conditions (financial, technical, social, emotional, etc.) of a creative conversation in which the participants will produce, discuss, explicate, filter and internalize in their practice an evolving collective memory. This second type of management is obviously much more subtle than the first, since it involves the sensitive concepts of shared views, relational familiarity, trust and incentives to creativity. Finally, actual knowledge cannot be separated from the consciousnesses in which it is reflected in the present, or from the individual learning processes it starts from and returns to. This subjective dimension of knowledge obviously cannot be “managed” by some outside authority like a thing or an objective situation. It belongs to the inner world, that is, to the desire to learn and share, to individuals’ work on themselves or their autonomous discipline. Having clarified these points, I will speak in familiar terms about “KM”; just as in general usage, people say the sun rises even though they know very well that it is the Earth that revolves.

The question of KM becomes more complicated when we consider the contemporary fashion of personal knowledge management (PKM)⁶.

4.2.1.2. The cycle of personal knowledge management

In the new ubiquitous digital environment – especially in social media – people are confronted with information flows so varied and abundant that they must learn to process them systematically. The complete cycle of PKM can be broken down into several distinct steps.

4.2.1.2.1. Attention management

People must first learn to control their attention: they therefore have to define their interests, order their priorities, identify their areas of effective competency and determine the knowledge and know-how they wish to acquire. Once all this has been properly clarified, PKM practitioners must strive to concentrate on their objectives without letting themselves be distracted by the multitude of information flows that cross the field of their consciousness. This should not prevent them from remaining open or from usefully placing their preferred objects of attention in the overall context that gives them meaning. They also have to be able to relate to people who have priorities different from theirs. The balance between openness and selectivity is a tricky exercise that must constantly be refined.

4.2.1.2.2. Choice of sources

Once we have set our priorities, we have to choose our sources of information. In contemporary social media, these sources are mainly other people. We thus need to spend time examining the information flows produced by others in order to choose those that best correspond to our objectives. We must also identify the institutions, businesses, research centers, networks and organizations of every kind that offer the information that is most relevant to us. It goes without saying that we can follow the choices made by people we trust and who share our interests, either automatically (collaborative recommendation systems are proliferating) or manually.

4.2.1.2.3. Collection, filtering, categorization and recording of information flows

The information flows from all sources identified must be aggregated or assembled in a single place so that they can be filtered in the most practical way. The collection tools can be RSS feeds from selected sites or blogs, colleagues, experts or institutions followed on Twitter or other social media, participation in online forums or various automatic alert systems. The choice of sources is the first form of filtering. But even feeds from our favorite sources have to be roughly evaluated and categorized in order to eliminate redundant information as quickly as possible. The information that is not eliminated must then be explicitly categorized (tag, comment, source name, etc.). Tags permit flexible, emergent categorization by means of freely chosen labels (social tagging) and the formation of networks for sharing references (for example, among researchers). Generally, only categorized information will be able to be used by others sharing the short-term collective memory (e.g. Twitter or Facebook) or long-term collective memory (e.g. YouTube, Flickr, Delicious or CiteUlike) where it is accumulated. It is impossible to classify without having a classification system, whether this system is implicit and unconscious or explicit and deliberately constructed. It is in our interest to make our own classification system explicit, if only to be able to perfect it and construct a more refined and effective memory.

4.2.1.2.4. Synthesis, sharing and conversation

Once information has been filtered, categorized and recorded, we need to be able to make a critical, creative synthesis. Only by so doing can we assimilate the information and transform it into personal knowledge. This synthesis, which as a rule is periodic, can be carried out in a blog, in an article, by editing a wiki entry, in a video, through incorporation into a computer program or in any other way. The essential point is to make the synthesis public, i.e. to introduce it into the open process of creative conversation of a community or network of people. The creative synthesis will be indicated in social media or disseminated through an RSS feed, or will feed an open source collaboration process or be made accessible through search engines and reported by automatic alert or recommendation systems or through the online social activity it generates. The synthesis will thus inevitably be exposed to criticism and comment from a community of people interested in the same subjects.

4.2.1.2.5. The feedback loop of personal knowledge management

In short, we pick up information, assemble it, categorize it, filter it, synthesize it, share the synthesis with others and then repeat this cycle creatively, always keeping a critical eye on our methods and tools. In this way, we prevent fossilization of our reflexes or blind attachment to our tools. After receiving feedback from creative conversation, we must periodically question our priorities, redefine the context, connect to new sources and eliminate old ones, perfect our filtering and classification tools, explore new methods of synthesis, get involved in other conversations, and so on. In doing this, PKM practitioners help not only themselves but also others to whom they are connected and who are doing the same thing.

4.2.1.2.6. Techniques pass but cognitive function remains

We must avoid unduly reifying the tools I have mentioned, which are only those used in the most advanced practices of 2011. In fact, in a few years, they will undoubtedly be replaced by new tools, or all aspects of PKM will be brought together in technical environments yet unknown as I write these lines, e.g. new types of browsers. In any case, the need for a personal discipline for collection, filtering and creative connection (among data, among people, and between people and data flows) will remain for a long time. Techniques pass but cognitive function remains. Without denying the importance of collective strategies and the shared visions that support them, I believe that social KM should be thought of as an emergent level based on the creative conversation of many individuals’ PKM. One of the most important functions of teaching, from elementary school to the different levels of university, will therefore be to encourage the sustainable growth of autonomous PKM capacities in students. This personal management should be conceived from the outset as the elementary process that makes the emergence of the distributed processes of collective intelligence possible and which in turn feed it.

4.2.2. The role of explication in social knowledge management

Let us make an inventory of the content of the memory of a knowledge community.

It is, first, all the signifiers recorded and manipulated by the community: these are documents in general, texts, images, sounds, multimodal signs, software, etc.

Second, we need to consider the languages or symbolic structures that organize signifieds and make it possible to read documents: jargon, classifications, thesauruses, codes, correspondences among various systems, etc.

Third, we need to add “abstract machines”⁷, ways of doing things, pragmatic rules by which documents are activated or processed, symbolic structures and relationships among people: methods, customs, know-how, and criteria and conventions of all kinds, which are often implicit. These rules include the methods of measurement, evaluation and judgment that produce the formally quantified or qualified data that are stored in the organization’s memory. Only mastery of these methods makes it possible to connect the documents to their referents.

Finally, we must consider a fourth aspect of the symbolic organization of a knowledge community that is not located at the same logical level as the others and ensures its self-referential looping. I am thinking here of reflexive reification, the work of self-modeling that allows the community to synthetically represent its own emergent cognitive processes to itself. We can say that one of the goals of KM is to support this self-referential modeling in such a way as to encourage the improvement of the processes of collective intelligence and facilitate individuals’ identification of their own roles (and those of others) in creating and maintaining the knowledge of the group they belong to.

Whether we are producing useful documents, clarifying or improving shared symbolic structures, spreading the most effective methods and practices or raising individual and collective awareness of the emergent cognition of the community, we will almost always find ourselves confronted with the problem of explicating implicit knowledge and processes.

The distinction between explicit knowledge and implicit knowledge echoes other dialectical pairs of opposites of the same type, such as objective knowledge and subjective familiarity or formal knowledge and practical competency. I suspect that the opposition between implicit knowledge and explicit knowledge in a new context reactivates the very ancient philosophical distinction between theoretical knowledge and empirical knowledge.

The explication of knowledge was studied and developed by the father of contemporary KM, Ikujiro Nonaka⁸. Nonaka proposed a cyclical model of the cognitive life of organizations. According to this model, called SECI (Socialization, Externalization, Combination, Internalization), knowledge exists first of all in an implicit form in individual practices. These practices are then socialized (S) and shared informally to become incorporated into organizational cultures. The critical phase of KM in organizations, according to Nonaka, is the transition from implicit knowledge to explicit knowledge (E). This externalization begins with a practice of questioning and dialog, which can only develop in an atmosphere of trust. It essentially consists of representing the largest possible part of the informal practices and the surrounding culture in the form of written documents, software or databases. The explication of knowledge has many advantages: it makes it possible to decontextualize and thus distribute and share information on a large scale, to critically examine the state of knowledge and possibly even to automate its application. The externalization of knowledge takes the form of explicit concepts, classifications or (computer) ontologies, methodological documents, rules, algorithms or programs. Once knowledge has been formalized in concepts and rules, it can be distributed in the information system of the organization, combined (C) and applied – possibly automatically – to the data flows that indicate the internal state or environment of the organization. The personal learning effort is not forgotten, since in the end the results of the explication and combination phases have to be integrated or internalized (I) by collaborators in order to be implemented, tested and perhaps transformed in practice. This will lead to a new cycle of socialization, questioning, dialog, formalization, recombination, and so forth. The organization’s knowledge is the life cycle I have broadly outlined, and not any one of its phases, artificially isolated. This model provides a general conceptual framework in which the organization can represent its own cognitive functioning to itself.

The SECI model was developed at a time when the Internet already existed but the Web was very new and social media were still unknown, except for a few pioneers of virtual communities. As I suggested above, our view of KM today draws much more on collaborative learning networks using social media than on the administration of central information systems controlled by experts. We need to promote organizational cultures and technical environments conducive to transparency, flexible reorganization of skill networks and continuous collaborative creation of immediately usable knowledge. Despite this, this dialectic of socialization, explication, combination and practical integration is still relevant for understanding the sustainable functioning of a creative conversation that produces knowledge.

The emergent discipline of KM has taught us that there can be no systematic exploitation of the knowledge capital of a community without the explicit modeling of the intellectual and social functioning of that knowledge capital. The following three points clarify the main relationship that in my view connects knowledge communities assembled around a common memory, on the one hand, and the models that explicate the functioning of their knowledge capital, on the other hand.

The first point I would like to make here is that we must not confuse knowledge capital with its explicit modeling. The map is not the territory⁹. A code of law does not encompass the living system of a nation’s mores. An English dictionary and grammar book provide only a “snapshot”, a partial image of a language spoken by a population dispersed over five continents and evolving in multiple forms. An explicit model is less than the living knowledge capital it reflects and disseminates. It is only an abstraction – and I would add, only one possible abstraction – of that reality.

My second point is in a way complementary to the first: there is no model that does not coproduce the reality it models. A map brings into being a territory where there are only experiences of movement and memories of travels¹⁰. Through its perlocutory force¹¹, a code of laws transforms the mores of a nation. Dictionaries and grammar books influence learning in school and the literary practices of languages¹². The model is a factor in the reality it explicates.

Third, the types of technical media used for the reflexive modeling of knowledge profoundly determine the identities of its referents. The old handwritten portolanos of medieval sailors, printed maps using the Mercator projection, dynamic online maps that combine GPS, satellite images, quick zoom-ins and zoom-outs on the screen of a laptop or an electronic tablet all structure our relationship to space and travel. Knowledge that is reflected in and transmitted through sung narratives does not have the same flavor as knowledge that is formalized logically in writing. And if this knowledge is represented in an online database and computer programs that automate reasoning, we are dealing with a third scenario that is different yet again. The medium of the model articulates not only the model itself, but also the distributed cognitive process that is modeled¹³.

To reproduce, improve and expand its shared memory, any social learning organization must have an explicit modeling method for the cycles of cognitive operations it carries out on data flows. It must create a (multimedia) image of the signifiers, systems of concepts and pragmatic rules that are part of its operations. Each of its participants must be able to filter, find, synthesize, analyze and comment on the data accumulated in its technical memory. One of the main effects of the explication of knowledge is that it makes its “distribution” beyond the geographic and social contexts in which it emerged possible. In short, knowledge must be reified and mediated so that it can be better shared. It can then benefit a broader community than the one (perhaps local or limited) where it emerged. Rather than knowledge being shut up in silos and Balkanized within small closed communities, one of the ideals of social KM is clearly its decompartmentalization, exchangeability and commensurability. An intelligent collectivity or a collaborative learning network has a truly shared memory only insofar as that memory is constructed and modeled by the creative conversation of its members in a unifying medium.

4.2.3. Dialectic of memory and creative conversation

Before going further into the question of the unifying symbolic medium of the memory, in order to make the reader realize its importance I would like to help elucidate the complex relationship between shared memory and creative conversation. To start with, where does the word conversation come from? Etymological dictionaries tell us that the verb to converse originally meant “to live with or among, to keep company with”. It was only in the 17th Century that it acquired the meaning of talking together or exchanging ideas. However, versare in Latin means “to turn or return”, and the prefix con- comes from the Latin cum, which means “with”. I am therefore proposing a hypothetical first etymology according to which, in con-versation, people turn to each other and exchange the direction of streams of discourse addressed to each other. According to my second hypothetical etymology, conversation is a process of con-version of knowledge from an implicit mode to an explicit mode and vice versa, and this reciprocal conversion is done “together” (cum).

Returning to the cosmic compass I have been using as an orientation instrument since the beginning of the chapter on the nature of information¹⁴, I would say that its intertropical zone is made up of processes of creative conversation. Its southern hemisphere consists of actual (implicit) processes of perception and action and its northern hemisphere is a virtual (explicit) memory shared online, removed from the flux of the immediate present. Creative conversation is thus the active interface, the original environment or source of the process of individuation of the knowledge community¹⁵. In the south–north direction, it transforms knowledge that is implicit, opaque, immersed in action, into shared virtual memory. In the north–south direction, it transforms the accumulated shared memory into actual effective sensory-motor activity.

Although physical meetings remain essential for establishing trust, increasingly conversational interactions oriented toward collaborative learning are taking place online, e.g. through social media. Judging by my personal experience on Twitter, the most constructive exchanges consist of short messages pointing to URLs containing multimedia data. The messages categorize these data with a brief comment and/or a hashtag¹⁶, a metadata label. Hashtags are used to bring together and find URLs, discussion threads or comments on a subject on specialized search engines¹⁷. The now increasingly widespread experience of watches or collaborative learning using social media makes it possible to observe in action how a creative conversation constructs a shared memory and is in turn constructed through the relationship to that memory. The immense flow of raw data is filtered and categorized by certain participants. Other participants confirm¹⁸ or dispute these categorizations, which may lead to discussion. The members evaluate the relevance and validity of the filtered data, reading recommendations and categorizations on the basis of their experience and knowledge of a field of practice¹⁹. If they are engaged in an active learning process, they will integrate the information received into their PKM systems, which in the end will transform their practice, and will also disseminate the information in other circles of conversation. The data are thus filtered, categorized and recategorized by a community, then found (by means of metadata) and used in practice by individuals, which changes the personal capacities of these individuals to filter and categorize, and the cycle begins again. This is how a conversation engine accumulates (data) and organizes (metadata) its shared memory. Through the integration of memory into practice and personal experience, creative conversation transforms data into knowledge. Symmetrically, implicit knowledge is transformed into data through blog entries, wikis and articles, and into metadata through an activity of participatory categorization.

The process of collaborative production of shared memory favors individual learning insofar as the individuals involve their personal experience in the conversations (the process of explication is always instructive) and involve the results of the conversations in the reorganization of their personal experiences. Here there is no purely individual learning, since data are exchanged and pooled. The imposition of metadata in a shared memory assumes a system of metadata common to a community. An open conversation validates the relevance of these metadata or diversifies the categorization of the data²⁰. There is no purely collective or only emergent learning, because the relevant filtering of data and the validity of metadata are ultimately based on experience and personal judgment.

We have seen that creative conversation organizes the dialectic of the relations between data and metadata. At a first degree of elaboration, the data – since they are externalized and shareable – belong to explicit knowledge. If we focus only on an analysis of digital memory, however, disregarding the living know-how, then the data belong to the implicit, opaque pole, while the metadata occupy the explicit pole that generates transparency and exchange. The explicit/implicit or virtual/actual polarity is thus more a matter of a pattern fractally repeated at various levels of analysis than of a clear and distinct separation between fields of being or knowledge. Thus, from the perspective of the constitution of shared online memory, creative conversations carry out an activity of “stitching” or interfacing between the opaque actuality of data flows (digitized phenomena, including texts) and the transparent virtuality of metadata (which make it possible to organize and search for information).

What do we call the characteristic site of this creative conversation that reciprocally converts virtual and actual modes of knowledge? Nonaka²¹ proposes that it be called ba, following recent developments in philosophy in Japan²². Ba is a place in the broadest sense of the word, that is, it can be material or institutional or based on a digital social medium. Its main characteristic is to enable the actual world of pragmatic action and the virtual world of discursivity to communicate within the same encompassing unit. From the point of view of social KM, ba is a condition of the creative conversation that feeds the life cycle of the knowledge of a collectivity. From the point of view of a more “emergentist” approach, we could say that ba springs from the creative conversation when a community succeeds in individuating (or in self-maintaining its process of individuation) around an activity of knowledge creation and sharing. In my view, in order to understand ba, it is best not to artificially separate the following three partial types of ba:

– the usual physical environments: offices, classrooms, meeting places;

– various digital environments: certain communities are organized using Facebook, LinkedIn or Ning groups and hashtags and subscription networks on Twitter, and networks on Delicious or Diigo;

– occasional encounters, such as conferences, symposia and seminars.

If all these times, places and social media are used by the same network of people, they become the components of a unique ba supporting the network’s knowledge creation process. It is creative conversation and its emotional tone that will unify all the communication and meeting media in a welcoming ba, and not any specific medium or architectural element labeled ba that will magically create a satisfying and productive knowledge community. In short, ba is the milieu associé, the environment specific to creative conversation, and it is being built as the knowledge community is individuated and its collective memory grows and is organized²³.

I note in conclusion that the collective individuation of a knowledge community is accompanied by processes of personal cognitive individuation on the part of its members. This personal cognitive individuation takes place horizontally, in social relationships of mutual aid, interactions among peers or relationships of users with discussion leaders of the community. Specifically, the type of effective participation by individuals in a community (rather than their official status or place in an organizational chart) will shape their social roles as experts, discussion leaders, collaborating learners or more passive users. Personal cognitive identity is also formed vertically, insofar as in each community individuals occupy specific semantic places according to their areas of expertise and learning paths. These places are identified by the traces the individuals leave through their activities of construction and use of the shared memory. While each knowledge community constitutes a distinct cognitive microworld, it is clear that the same areas of personal expertise will be projected differently in different communities. It should be noted in this regard that the names of users or persons often serve as markers of semantic zones. In many social media, in fact, subscription to a feed from a particular user may be interpreted as a statement of interest in the subject in which the user specializes²⁴.

In short, creative conversation transforms implicit personal and local know-how into explicit knowledge codified in a collective memory. This construction of a shared memory implies distributed work of production, filtering, categorization and evaluation of data. In its dimension of personal integration or learning, creative conversation in turn transforms explicit knowledge into know-how applied locally in the corresponding fields of practices. This alternating cyclical transformation is coordinated in a milieu associé, ba, which cuts across and unites the organizational mechanisms, physical places and digital environments that support the conversation. Finally, creative conversation is the source of personal and collective processes of cognitive individuation that determine its consistency and duration.

4.3. The symbolic medium of creative conversation

4.3.1. The question of the symbolic medium

The preceding descriptive analysis, which deals with the ideal creative conversation, could leave the impression that all is for the best in the best of all possible digitized worlds. But this is not the case. In fact, we are currently a long way from possessing the symbolic medium – or the intellectual technologies derived from that medium – that would allow us to obtain the greatest advantage from the distributed creative conversations whose memories are accumulated on the Web. The problem is threefold. It has to do with the transversality of individuals with respect to communities, the transversality of communities with respect to digital environments, and the transversality of knowledge with respect to the various memories accumulated by communities.

First, a single person usually participates in several social or occupational networks, or various knowledge communities. Individuals thus act as “crosspollinators” among various cognitive ecosystems. Communities use different languages, modes of conceptualization and metadata systems. The problem arises because a personal knowledge management system should be able to automatically²⁵ be fed information and in turn feed the online memories of the knowledge communities the person takes part in. Today we are a very long way from that. The data formats of these memories are often incompatible²⁶, and their metadata systems (the conceptual organization or classification) even more so. In addition, the general view is that automatic language translation systems work well enough to provide a quick idea of the content of a text or the meaning of a word, but that they cannot be used to transfer information from one language to another reliably – and acceptably in terms of reading quality – without serious human revision. In fact, few French, American or Brazilian Internet users have any idea of the content of the Chinese blogosphere or the Japanese Twittosphere, and vice versa.

Second, a single knowledge community often uses many applications and digital environments, as I stated in my discussion of ba above. For example, a college or university class may use Delicious, as well as both Facebook and Twitter groups, while a community of professionals may use a LinkedIn forum, Diigo, a network of blogs, etc. We encounter the same problems as those mentioned above regarding people’s participation in many different knowledge communities. It should be noted, however, that interoperability among various services supporting creative conversations is developing, thanks to the spread of open APIs²⁷ and third-party applications specializing in data transfer. To give two simple examples: when I post a message on Twitter, it is reproduced in my Facebook, Friendfeed, LinkedIn, etc., feeds, and when I bookmark a page on Delicious, the URL is indicated in my feeds on Friendfeed, Facebook, Plaxo, etc. We are still far from having transparent circulation among online knowledge management applications or eliminating barriers among competing social media, however, particularly in terms of the semantics of categorization processes.

Third, there are obviously many communities that should be able to connect their memories, especially when all or parts of these memories concern the same subjects. Despite this, once again the disparate nature of classifications and metadata systems, not to mention the multiplicity of languages, makes such connections, or even the suggestion of them, difficult to automate.

Knowledge management on the Web is still too collectivized, in fact Balkanized among many competing services, languages and ontologies. The situation is often much worse in big companies and public administrations, whose databases are frequently unable to communicate with each other. With the possible exception of blogs, paradoxically, most PKM tools are centralized by big companies specializing in social media and search engines. Just as computer science underwent a revolution in the 1980s with the widespread use of personal computers, it is possible that KM in the 21st Century will experience a decentralizing revolution that gives more power and autonomy to individuals and self-organized groups. This can only take place through the adoption of a common protocol for the expression of semantic metadata, which would free creative conversation from the limits imposed by the major players of the Web²⁸. Through such a semantic protocol, operating as a shared tool for explication and modeling, creative conversation could fully realize all its transversal potential: people participating easily in many communities, communities transparently using many applications, and information being exchanged and connected automatically among the memories of various communities. Above all, the adoption of a shared semantic metalanguage would make it possible to advance toward a social KM that would emerge without too much friction from autonomous practices in PKM, and that would ultimately serve these practices. We thus come back to the question of a unifying symbolic medium, with which I ended the section on the role of explication in KM²⁹.While the Internet is currently the unifying medium in terms of techniques for the material communication of messages, we still do not have a symbolic medium or common language that allows us to share knowledge in a computable and transparent way and thus to develop a creative conversation on a global scale, with all the resulting benefits we can expect in terms of human development. It is only on condition that such a symbolic medium exists that we will be able to properly speak of online explicit knowledge as a commons³⁰ that is actually usable by everyone according to the goals and viewpoints of all communities.

4.3.2. The metalinguistic articulation of organized memory

The question of how to organize recorded information in a coherent and useful memory is not new. In fact, it is as ancient as libraries. In the 17th Century, when the proliferation of print publications led to a huge increase in the number and size of libraries, the problem of how to classify publications became very urgent. Those responsible dealt with this problem of organization by proposing a metalinguistic articulation, just as I am doing today. Since it is not advisable to imagine the future without recognizing the heritage of the past, I would like to provide a broad outline of the main stages of thought on document metalanguage, associating each of them with a “big name”. What follows is not an exhaustive history of the documentation sciences, but merely the identification of some of their main paradigms³¹.

At the end of the 17th Century, the philosopher and mathematician Leibniz studied the writing of the 14th-Century Dominican kabbalist Ramon Llull on the art of mechanically producing true propositions³². He took an interest in the ideography and hexagrams of the I Ching³³, which the Jesuits had just brought back from China. He explored binary arithmetic and combinatorics. He built the first calculating machine capable of performing the four arithmetic operations. Bringing together all these areas of practice and thought, he imagined a writing system he called the universal characteristic, which would be able to express and combine all ideas mathematically. Leibniz’s work had a strong influence on the founders of contemporary logic and computer science³⁴. Leibniz was a librarian for 40 years, and thus had to deal with the concrete problems of managing the catalogs of many libraries. He is the first philosopher and scientist to think rigorously about the problem of classifying knowledge as it relates to the organization of libraries³⁵. He theorized on the need for a metalinguistic layer that would be distinct from the documents and would help users find their way around the library: abstracts and indexing. He also imagined the ideal physical architecture for a library, reflecting the organization of knowledge: a kind of panopticon of knowledge. In one of his works, he even tried to calculate the maximum size of a future universal library of humanity³⁶.

In the 19th Century, the American Melvil Dewey created the decimal classification system that bears his name³⁷. The decimal classification is rational and universal, and independent of institutions, languages and physical establishments. An advance compared to the systems then in use – which assigned books to certain shelves – Dewey’s system provides a hierarchical (nested categories and subcategories), exclusive (a document cannot belong to two separate categories) classification of knowledge. Most classification and indexing systems in use today – including the American Library of Congress system and the French RAMEAU system³⁸ (itself based on the Library of Congress system) – are derived from the hierarchical classification created by Dewey, although they are both more flexible and more complex.

In the 20th Century, the Indian mathematician Ranganathan, one of the founders of modern documentation sciences, restructured the profession of librarian around users³⁹, called for a universal semantics and invented a faceted classification system⁴⁰. The system was based on the principle of intersecting categories – or the composition of “semantic primitives” and allowed a document to be found from several perspectives. The metalanguage created by Ranganathan (Colon Classification) has been used very little outside India, but the principle of faceted classification was accepted and frequently applied in other forms.

Even before the Second World War, the Belgian Paul Otlet had considered the theoretical problems of a universal library and its indexing. Otlet popularized the microfiche – which was already in use in the USA – in Europe. He created the Universal Decimal Classification (UDC), an adaptation of the Dewey system that was more flexible⁴¹ and used faceted language. He then undertook an unfinished project of building a collective memory of humanity for the League of Nations. In his books Traité de Documentation (1934)⁴² and Monde (1936)⁴³, Otlet conceived of a networked intellectual world coordinated by means of a classification system that was universal but would constantly be reconfigured according to links created among documents by users. This was the first detailed formulation of the principle of hypertext interconnection, before those of Vannevar Bush, Douglas Engelbart and Ted Nelson⁴⁴. Paul Otlet had a coherent vision of the world of documents as a growing ecosystem and he foresaw that electronic technologies would soon make information ubiquitous (he was writing in the 1930s!).

4.3.3. How can creative conversation organize digital memory?

The classification and indexing systems that allow library users to find the documents they are looking for work well. Why not use them on the Web? Why invent a new metalanguage when so many already exist and have proven their worth?

Software forms of memory are very dependent on their material and technical media. The indexing methods and document metalanguages developed and perfected in the 19th and 20th Centuries were designed to manage searches for print documents or material media in physical institutions or, at most, national networks of institutions. The existence of a large number of different classification and indexing systems in the world did not create too many problems as long as each library or documentation center was organized using a single system. However, since the beginning of the 21st Century, practically all libraries, museums and archives have been digitizing and offering not only their catalogs but their collections online. As a result, human memory tends to be collected in a single technical medium. Consequently, national and institutional disparities in indexing and classification methods or document metalanguages are no longer acceptable in the long term.

This is one of the reasons library and documentation sciences have been undergoing a major reexamination since the public appearance of the World Wide Web around the end of 1993⁴⁵. The size of the memory has grown immeasurably: a universal multilingual multimedia library is on the horizon. The documents and the links among them are undergoing constant change, being almost fluid. The general interconnectedness and ubiquity are changing users’ search methods and practices. If we analyze the current situation through the eyes of future generations, it is clear that the possibilities for automatic calculation and interactivity are still largely underused for lack of standards and metalanguages suited to the new conditions.

We need to develop new ways of thinking about archives and their organization in order to deal with the elimination of constraints involving the physical location of documents – constraints that have existed since the beginning of writing 5,000 years ago. In fact, all documentary systems and indexing metalanguages prior to the Web have had to deal with the eminently practical imperative of the material placement of documents. The need to store information media “somewhere” seemed so natural that it was hardly recognized as a real constraint. As David Weinberger points out⁴⁶, it was not only the library’s books, disks and cassettes, but even the files and catalogs, that required three-dimensional spatial organization. Since the existence of the Web – a very recent phenomenon on the scale of cultural evolution – digitized information has proliferated and it can be distributed indefinitely to every node on the network at minimal cost. Archives can be multiplied at will or reached by pointing to hyperlinks in the ubiquitous (i.e. ever present) digital environment; they thus no longer first have an address in physical space⁴⁷, but in an intangible semantic sphere. It is their meaning or relevance to readers that now constitutes their main address. The basic addressing has gone from the physical order (the library call number) to the semantic. This change leads to a second one: the possibility of indefinitely varying the semantic addressing of a document according to points of view and uses. As I pointed out above, and as actors/users of the participatory Web know from experience, it is now possible to structure and index the same set of documents in a thousand different ways. It is no longer only experts in documentation and information sciences, using well-established methods, who classify documents, but billions of users, tagging them in their own ways⁴⁸. Indexing, until recently reserved for experts, is now practiced on a large scale by anyone and everyone on Amazon, LibraryThing or YouTube⁴⁹, social bookmarking sites, blogs, Twitter and, thanks to Faviki, even Wikipedia. The result of this collective classification activity is called a folksonomy (the word is modeled on taxonomy). It is true that the tags of folksonomies are inconsistent because of synonymy (many key words are used to designate the same concept) and homonymy (some key words have many meanings), not to mention the noise introduced through spelling mistakes, plurals, abbreviations, etc. In addition, the tags correspond to very disparate levels of generality and cannot readily be organized in classes and subclasses. Finally, the multiplicity of natural languages (in which the tags are usually expressed) still seriously fragments the creative conversations that have been starting in the last few years to organize the global memory. As imperfect as the folksonomies of 2010 are, however, they prefigure the creative conversation of the future, which will be capable of providing as many points of view for the universal memory as there are human communities and interests.

This perspective allows us to glimpse an emerging new type of metalanguage, a kind of writing in the second degree. This “meta” writing no longer places – or no longer only places – signs on a page or even on a screen, but attaches them to flows of digital data. Of course, as we have just seen, the concept of a document metalanguage is very old, but I am speaking here of a new generation of metalanguage: universal, democratic and calculable. This new language will be universal because memory is now world-wide. Unlike previous metalanguages, which were all local and based on a single culture, the new metalanguage will have to be radically equanimous, capable of expressing the perspective of any culture or tradition⁵⁰. It will be democratic because its manipulation will no longer be the preserve of information experts, but open to all participants in creative conversation by means of sensory-motor interfaces and translation into natural languages. Finally, it will be calculable, because all previous metalanguages were designed before the digital medium and its almost unlimited calculating power. The new metalanguage will make it possible to categorize information, evaluate it according to different rules, and trace navigation routes through the ocean of data⁵¹. Semantic computation based on the new metalanguage will not be limited to automated reasoning that infers the properties of a class from its belonging to a super-class. It will be able to generate and regenerate at will the hypercomplex fractaloid graph of formal concepts that will encompass the huge mass of information in their regular net. Obedient to the billions of pairs of hands in the creative conversation, this new kind of computation will steer the trajectories of attention and value in the unlimited semantic sphere that coordinates the library of Babel⁵². To transform the deluge of information into useful, organized memory carrying knowledge across language barriers, moving with ease through the diversity of cultures, the creative conversation that arises from cyberspace needs a symbolic medium in keeping with its scope.

---

1 See [LÉV 1997].

2 For how the new digital mediasphere can enrich the democratic process, particularly public deliberation, see my two books Collective Intelligence [LÉV 1997] and Cyberdémocratie [LÉV 2002]. See also Manuel Castells’, Communication Power, Oxford University Press, 2009 [CAS 2009]

3 See The Wisdom of Crowds, by James Surowiecki [SUR 2004] for a recent discussion of this subject. See also “Economics and knowledge” [HAY 1937] and Law, Legislation and Liberty [HAY 1979] by Friedrich Hayek. Hayek was one of the first to provide an explicit theory of the emergence of a spontaneous order based on interaction among responsible individual intelligences. This spontaneous order is obviously not perfect for any one person, but it is generally better than an order planned by a small group of leaders, because it incorporates distributed knowledge of the complexity of real situations, knowledge that is more accurate, rich and varied. I have dealt with the subject of competitive cooperation in the economy and elsewhere in my book World Philosophie [LÉV 2000].

4 See, for example, Convergence Culture, by Henri Jenkins [JEN 2006], which clearly demonstrates the collective intelligence of communities of fans, displayed in online creative conversations.

5 For a general overview of the field, see Kimiz Dalkir, Knowledge Management in Theory and Practice [DAL 2005].

6 See, for example, “Personal knowledge management: putting the ‘person’ back into the knowledge equation”, by David Pauleen [PAU 2009]. It is clear that PKM is not a contemporary invention: only the conditions and tools are new.

7 I borrow the term from Deleuze and Guattari in A Thousand Plateaus [DEL 1987b].

8 The pioneering work, already quoted in the introduction of this book, is The Knowledge- Creating Company: How Japanese Companies Create the Dynamics of Innovation [NON 1995]. See also Enabling Knowledge Creation: How to Unlock the Mystery of Tacit Knowledge and Release the Power of Innovation [NON 2000].

9 See Alfred Korzybski, Science and Sanity, An Introduction to Non-Aristotelian Systems and General Semantics [KOR 1933].

10 On this point, see Bruno Latour, “Les vues de l’esprit, une introduction à l’anthropologie des sciences et des techniques” [LAT 1985].

11 On the concept of the perlocutory force of performative statements, see John L. Austin, How to do Things with Words [AUS 1962].

12 This point was emphasized by Sylvain Auroux in La Révolution Technologique de la Grammatisation [AUR 1994].

13 The role of the communications media in symbolic organization will not be discussed in detail in this chapter. Among the huge mass of scholarly work on this subject, I will mention only works by McLuhan [MAC 1962, MAC 1964] and myself [LÉV 1990, LÉV 1994b, LÉV 1997].

14 See section 2.1.

15 On the concept of individuation, see Gilbert Simondon, L’Individuation à la Lumière des Notions de Forme et d’Information [SIM 1958a].

16 A hashtag is a keyword preceded by a hash symbol (#), e.g. “#PKM” to indicate that the “tweet” (the message and the URL it points to) concerns PKM.

17 For example, Twitter search, Twazzup or Topsy (in 2010).

18 A mark of confirmation on Twitter is re-twitting messages considered most relevant, that is, forwarding them to your own subscribers.

19 In emphasizing the importance of a shared practice (at various levels of expertise), which needs to be combined with a community of people and a common subject to obtain a creative conversation, I am in agreement with Etienne Wenger’s studies of communities of practice. See his Communities of Practice: Learning, Meaning, and Identity [WEN 1998].

20 Contrary to what happens, for example, in traditional libraries, it is always possible to categorize the same document in many ways, according to the various points of view of the users. For more information on the freedom of open categorization through collaborative online memories, see David Weinberger, Everything is Miscellaneous: The Power of the New Digital Disorder [WEI 2007] and the online article by Clay Shirky “Ontology is overrated” [SHI 2005].

21 For example, in his article “The concept of Ba: Building a foundation for knowledge creation” [NON 1998] and his book Enabling Knowledge Creation [NON 2000].

22 See K. Nishida, Fundamental Problems of Philosophy: The World of Action and the Dialectical World [NIS 1970] and An Inquiry into the Good [NIS 1990]; specifically on the question of the creation of information, see H. Shimizu, “Ba-Principle: New logic for the realtime emergence of information” [SHI 1995].

23 On the concept of the milieu associé, see Gilbert Simondon, L’Individuation à la Lumière des Notions de Formes et d’Information [SIM 1958a]. The processes of individuation obviously have a counterpart in processes of dissolution: communities are not eternal.

24 Etienne Wenger stresses the importance of the construction of identities in communities of practice; see his book, cited above, on communities of practice [WEN 1998]. My work on knowledge trees [LÉV 1992a] also presents – and graphically models – this relationship of reciprocal construction of personal identities and collective identities in online knowledge communities.

25 This automation includes filtering controlled by individuals as well as collaborative filtering that selects information according to its relevance for a group of people whose choices are similar.

26 The increasing adoption of the XML standard and, with more difficulty, the RDF standard (both proposed by the WWW consortium), as well as the use of other data exchange formats such as JSON should in principle make it possible – eventually – to overcome the obstacle of the incompatibility of data formats.

27 API stands for Application Programming Interface, an interface that can be used by a program external to a particular service. These interfaces facilitate data transfer and form the basis for interoperability between services.

28 I am not speaking here of a protocol on data or metadata formats – this work is being pursued today by the WWW consortium and other standardization organizations – but of a symbolic system, a language in the full sense of the word, such as IEML, which is especially designed for semantic calculations and interconnections.

29 See section 4.2.2.

30 I will discuss the subject of the commons below. See also section 6.1.2.

31 For an overview of the intellectual principles of the documentation sciences, see Elaine Svenonius, The Intellectual Foundation of Information Organization [SVE 2000].

32 The main reference on Llull’s work on logico-linguistic combinatorics is the Ars Magna [LLU 1990].

33 I Ching: The Book of Changes [YIJ 2002].

34 On Leibniz’s thought, see Gilles Deleuze, Fold: Leibniz and the Baroque [DEL 1993]; Michel Serres, The System of Leibniz [SER 2003]; Yvon Belaval, Leibniz Critique de Descartes [BEL 1960]. In the introduction to his first book on cybernetics, Norbert Wiener outlines what the new science of computers owes to Leibniz’s thought: to explain a fact, it starts with a matrix of possibilities and then tries to understand why one particular possibility was realized rather than another, whereas Cartesian thought looks for real sequences of causality leading to the fact that is being explained. See Cybernetics [WIE 1948].

35 See Jacques Messier, “Un bibliothécaire parmi les humanistes: Gottfried Wilhelm Leibniz (1646–1716)” [MES 2007].

36 Gottfried Wilhelm Leibniz, De l’Horizon de la Doctrine Humaine, 1693 [LEI 1693].

37 A Classification and Subject Index for Cataloguing and Arranging the Books and Pamphlets of a Library, 1876 [DEW 1876].

38 RAMEAU (Répertoire d’Autorité Matière Encyclopédique et Alphabétique Unifié) is a language used to index the collections of public libraries.

39 See his Five Laws of Library Science [RAN 1931]. The five laws are: “(1) Books are for use. (2) Every reader his [or her] book. (3) Every book its reader. (4) Save the time of the reader. (5) The library is a growing organism”.

40 See his book presenting the principle of faceted classification: Colon Classification [RAN 1933].

41 The UDC is still in use, with nearly 65,000 subdivisions. See http://www.udcc.org/.

42 Nearly unobtainable until recently republished by the Centre de Lecture Publique de la Communauté Française de Belgique [OTL 1934].

43 See [OTL 1936].

44 See Vannevar Bush, “As we may think” [BUS 1945]. The work of Douglas Engelbart at the Stanford Research Institute in the 1960s has been documented by Thierry Bardini in Bootstrapping, Coevolution, and the Origins of Personal Computing [ENG 1962], [BAR 2000]. For Ted Nelson, see his Literary Machines [NEL 1980], several previous versions of which were published in the 1970s.

45 Although Tim Berners-Lee’s initial idea was published in an internal memo at CERN in 1989, there were still only 50 Web servers in the world in 1991. It was only with the first version of Mosaic by Marc Andreessen in September 1993 (which became Netscape in 1994) that the Web began to experience world-wide success.

46 In Everything Is Miscellaneous: The Power of the New Digital Disorder [WEI 2007].

47 Obviously, digital files still have to be located someplace in the physical memory of one or more servers.

48 See Isabella Peters, Folksonomies: Indexing and Retrieval in the Web 2.0 [PET 2009] and Gene Smith, Tagging: People-powered Metadata for the Social Web [SMI 2007].

49 For further information on YouTube as a medium of participatory cultural practices, see [BUR 2009].

50 To illustrate the narrowly ethnocentric nature of traditional document metalanguages, the following are the 10 subdivisions of category 200 (religion) of the Dewey Decimal Classification, one of the most widely used in the world:
200 Religion / 210 Philosophy and theory of religion / 220 Bible / 230 Christianity, Christian theology / 240 Christian moral and devotional theology / 250 Christian orders and local church / 260 Social and ecclesiastical theology / 270 Christian Church history / 280 Christian denominations and sects / 290 Other and comparative religions. If we wanted further confirmation of the ethnocentrism and dated nature of the Dewey classification, the following are the subdivisions of category 290 (Other religions): 291 Comparative religion / 292 Classical (Greek and Roman) religion / 293 Germanic religion / 294 Religions of Indic origin / 295 Zoroastrianism (Mazdaism, Parseeism) / 296 Judaism / 297 Islam, Bábism and Baha’i Faith / 298 Mormonisn / 299 Other religions.
It should be noted, for example, that Buddhism is not even mentioned directly and that the Baha’i faith – for which I have the greatest respect, but which has seven million members and whose followers are persecuted in many Moslem countries because it is not one of the religions of the book mentioned in the Koran – is put in the same category and on the same level as Islam, which has a billion and a half believers. We find the same absence of equanimity, the same ethnocentric myopia and the same dated quality of the classification in other areas of knowledge. Other classification systems (including the Library of Congress system, which is obviously dependent on the particular situation in the United States) are not much better in this regard. That is why, rather than a classification or super-ontology, I am proposing a formal language of creative conversation that will make it possible to express any concept and any classification.

51 Vannevar Bush spoke of creating lasting “trails” in the forest of the future computerized memory [BUS 1945].

52 See Borges’s famous story entitled “The library of Babel” [BOR 1998c].