Chapter 13

Hermeneutic Memory

13.1. Toward a semantic organization of memory

The Hypercortex can be seen as a memory, since it contains and organizes data. It functions as a hermeneutic memory that allows the application of multiple interpretation strategies. As we will see in this chapter, this memory is perspectivist – it integrates many distinct points of view – and is structured in layers of increasing complexity (data, information, knowledge). Different creative conversations can generate automatable functions in it as they see fit. Some of these functions will categorize and evaluate data, thus producing semantic information units, which are formal representations of ideas. Other functions will situate these information units in encompassing theoretical or narrative contexts that will specify or transform their meaning. The combinations of these hermeneutic functions form a multitude of collective interpretation games. It is thanks to the existence of a common metalanguage for computable semantics (IEML), which, as we have seen, provides a system of coordinates for the world of ideas, that all these games can converse and exchange their cognitive resources in the same general semantic information economy. The Hypercortex will thus function as a collaborative tool for enhancing individual and social knowledge management¹.

Figure 13.1 shows the location of Chapter 13 in the discussion of the general structure of reflexive collective intelligence. Neither the semantic machine nor the IEML metalanguage is a goal in itself. The main goal of the techno-symbolic system based on the IEML semantic sphere is the modeling or simulation of the world of ideas that reflects collective intelligence. It is precisely this monadological model of the world of ideas that will be described in this chapter on the hermeneutic memory of the Hypercortex.

13.1.1. Implications of collective processes of categorization in the digital medium

For the first time in history, thanks to the digital medium, humanity is cultivating an interconnected shared memory in which ubiquitous data can be transformed at will by symbol-manipulating automata. The public has only had access to the Web since the mid-1990s, so the techno-cultural exploration of this shared memory has only just begun. Even before the appearance of the Web, a few intellectual technologies using computers had already emerged, such as spreadsheets, multimodal interactive simulations and hypermedia. My hypothesis, however, is that the major developments in the full symbolic and cognitive exploitation of a global digital memory are still to come. No generation before ours has been able to organize and use a practically inexhaustible flow of data of such cultural variety, produced by human communities present and past. To meet this challenge, we have to deal with the problem of augmenting our collective capacity to categorize and evaluate digital data.

Generally speaking, the activity of categorization is the key to cognitive processes, and this is particularly the case for human cognitive processes, which are organized in symbolic systems that are cultural in origin². More specifically, recent work has shown that the design and management of databases – which are ultimately organized by systems of categorization – is becoming one of the main scientific activities³ and perhaps the essence of digital art⁴. Social media such as Diigo, Facebook, Twitter, LinkedIn, Flickr and YouTube, as well as blogs, ask their users to actively participate in categorizing data. In the future, personal cloud management – managing personal data and services via the Internet⁵ – and online personal knowledge management will become widespread and systematic. Here again, the implications of the methods of categorization are central and they have impacts on the collaborative management of the knowledge that is taking shape in public administrations, corporations and research networks⁶.

The problems involved in categorizing data have become hypercomplex and gigantic in scale. The first efforts to solve these problems while respecting their complexity are beginning to see the light of day. I mention in particular research on improving processes of social tagging⁷ online and the web of data based on the RDF standard and ontologies expressed in OWL⁸.

13.1.2. A renewed approach to the problem of categorization

The reader who has reached this point is already familiar with my thesis: natural languages, like systems of notation invented before 2000, are not suitable for the nature and scale of the collective processes of categorization that will become current in the digital medium in the 21st century and beyond. Neither natural languages nor the traditional documentary languages (such as those used in libraries) were designed to exploit the new interconnected global memory and its calculating power. Natural languages are in harmony with the functioning of the human brain and are obviously not made to be processed automatically. The old systems of notation and writing were invented in times when methods of physical storage and retrieval of information were heavy, slow, manual and local, as opposed to the automatic, ubiquitous, ultra-rapid systems we have today.

Most search engines circumvent the problem of precise semantic representation of data and their free interpretation by users by carrying out statistical calculations on chains of characters – the primary purpose of which is to represent sounds (and not concepts) – or links – the primary purpose of which is to point to data (and not to categorize them).

Ontologies deal with the problem of categorizing data on the Web by constructing rigid logical relationships among chains of characters, most often URLs (which are semantically opaque). This is the heritage of expert systems and research on artificial intelligence (AI) that predate the Web, however, and these ontologies only formalize limited conceptual universes. That is why the web of data produces a large number of ontologies that are often disconnected from each other, while all digital data are potentially interconnected in the global memory.

In my view, trying to synchronize and optimize processes of categorization as varied and massive as those in the digital medium using natural languages and systems of categorization that pre-existed the Web is like trying to perfect algorithms for manipulating numbers encoded in Roman numerals instead of first seeking a better symbolic system for encoding numbers. We need a symbolic system for the notation and manipulation of concepts that is designed from the outset for massively distributed real-time social computing in an interconnected global memory. This is precisely what IEML is. In order for IEML to help each creative conversation to organize the digital memory for its own purposes while maintaining semantic interoperability with all the others, the use of the metalanguage must be based on a coherent philosophy of memory.

13.2. The layers of complexity of memory

We are familiar with the famous quotation from T.S. Eliot: “Where is the life we have lost in living? Where is the wisdom we have lost in knowledge? Where is the knowledge we have lost in information?”⁹. This poetic cascade of questions elegantly describes our situation in the practical world. It has been used in the world of corporate management for some 30 years¹⁰ to describe the chain of successive abstractions – or the scale of growing value – linking data (which we do not find in the Eliot quotation), information, knowledge and wisdom. Usually when the data– information–knowledge–wisdom (DIKW) hierarchy is invoked, it is to prevent confusion between one link in the chain and the next. Data (numerical or quantitative) are rudimentary, while information (readable and interpretable) is categorized, evaluated and commented on.

Information is isolated, however, whereas knowledge links information in patterns and gives it meaning in a context. This knowledge (of experts?) may eventually be explicated and formalized in a communicable theory, while wisdom (of leaders?) implies both profound experience in human affairs, the humility that comes from self-knowledge, and direct intuition that is impossible to express in a formula. The DIKW chain also suggests a process of progressive refinement or transformation, going from raw material, represented by data, to the most valuable but probably the most volatile final product – wisdom. Although the DIKW chain or pyramid has been cited for decades and has appeared in thousands of PowerPoint presentations, its elements have rarely been precisely defined. We would search in vain in scientific journals for a functional formalization of the transformations from one to the next.

I will now use the structure provided by this commonplace chain to describe the four degrees of complexity of memory in the cognitive model of the Hypercortex¹¹. I will propose my definition of each level of complexity as well as an operational approach to the transitions from one level to another.

I would like to begin by defining the hermeneutic approach to symbolic cognition, which will permit me to functionally formalize the DIKW pyramid.

13.3. Radical hermeneutics

13.3.1. Introduction to the hermeneutic approach to cognition

At its origin, hermeneutics was the art of interpretation of texts¹². Interpreting a text essentially means reading it in a deep, systematic way in order to extract the meaning that is most interesting and useful from the point of view of the interpreter and the community to which he or she belongs. This work of reading consists, in practice, of writing texts (peritexts, epitexts, paratexts of all kinds) about the text being interpreted. Reading and writing involve the same basic hermeneutic operations, in which the creation of meaning is central.

The concept of text in contemporary hermeneutics is much broader than that of the written notation of speech. On this point, I am drawing on postmodern and deconstructionist (in the broad sense) perspectives pioneered by authors such as Wittgenstein¹³, Foucault¹⁴, Derrida¹⁵ and Lyotard¹⁶, in which all types of symbolic arrangements can be considered texts, including those based on oral traditions or iconic, musical, ritual or other non-linguistic symbolic systems. In this approach, the processes of reading and writing texts involve an infinitely open multitude of language games, epistemes or functions of textual production and interpretation, no one of which is favored in principle. Ultimately, there is no objective, exterior or neutral meaning. Meaning is always produced by a particular mechanism of textual interpretation, and all hermeneutic (meaning-producing) machinery is necessarily dated and situated.

My conception of hermeneutics is also in keeping with the view popularized by Gadamer¹⁷, Ricoeur¹⁸, Gusdorf, Vattimo¹⁹, Grondin²⁰, etc., that the hermeneutic process in the human sciences, rather than producing an “objective truth” through the application of formal methods to cultural artifacts, translates a truth of human experience that is always singular and is embedded in a particular history. This understanding of hermeneutic activity has obvious affinities with Nietzsche’s perspectivism²¹, which must be distinguished from the view that “everything is equally valid”, which is typical of extreme relativism. With the freedom to interpret comes the responsibility to guide a community’s symbolic cognition, which requires great prudence²². My approach to hermeneutics also owes a lot to the concept of radical imagination developed by Cornelius Castoriadis²³, for whom human creations of meaning cannot be entirely reduced to some mechanistic determinism but derive from an autonomous, irreducible creative power.

13.3.2. The thesis of radical hermeneutics

My hypercortical model of cognition is consistent with a thesis held today by many scholars in faculties of arts and social sciences, which may be called radical hermeneutics. This thesis can be articulated in two interdependent propositions: first, it is impossible to separate symbolic cognition from memory and, second, all organization of memory is interpretative in nature.

13.3.2.1. It is impossible to separate cognition from memory

The physical recording of data (like the material capacity to store, classify, sort and organize data) is certainly a condition of memory, but it cannot be identified with memory itself. Common sense tells us that an indefinite accumulation of data without any form of organization does not make a very useful memory. We can think of a random pile of books as opposed to a library, where the books are arranged on shelves according to call numbers and can be found by author, title and subject in a catalog. Storehouses of data become real memories only insofar as they permit operations of selection according to semantic criteria (categorization: what is the document about?) and criteria of importance or relevance (evaluation: what is the value of the document?), which depend ultimately on an emotional investment. In addition, we know that data only become meaningful when structured by theories, narratives or other organizing perspectives. Memory therefore implies most other cognitive operations, in particular affective investment, categorization and discursive organization.

Similarly, it is difficult to conceive of any cognitive operation (perception, learning, problem solving, symbolic manipulation in general) that does not call upon short-term or long-term memory. How, for example, could a person understand an utterance if he or she only remembered the end of it and not the beginning (shortterm memory) or had no knowledge of the context shared with the speaker (longterm memory)? The same is true of music: without memory, we could not perceive rhythm or melody, and music itself can bring back memories. For these reasons we must not think of memory as something that can be separated from cognition, but as its temporal dimension, and it is impossible to remove cognition from time. This is true not only for personal cognition but also for social cognition: institutions, like cultures, necessarily function on the basis of a memory: archives, narratives, rituals, transmission of memories or traditions, landscapes shaped by humans.

13.3.2.2. All organization of memory is interpretative in nature

As we have seen, memory implies categorization, evaluation and some kind of narrative or theoretical ordering. All these operations result from interpretative choices – they are hermeneutic operations – because there are always other possible ways in which to organize data. Since all symbolic cognition implies the participation of memory and all memory results from interpretative choices, symbolic cognition is intrinsically hermeneutic in nature. In other words, according to the thesis of radical hermeneutics I am defending here, a cognitive system is an interpretive machine. This thesis can give rise to certain misunderstandings, which I would like to dissipate immediately.

13.3.3. Radical hermeneutics beyond the misunderstandings

First, interpretation is not necessarily a vague, “purely subjective” process. It is clearly possible to establish very strict rules of interpretation, as is done, for example, in linguistics, philology, law, theology, etc. The modern natural sciences are based on rigorous interpretation procedures. Scientific research does not preclude interpretation; it simply advocates methods of interpretation that are explicit, shared and unambiguous, that lead to observable predictions, etc. It is a commonplace in the epistemology of science that observable data have meaning only according to a theory, i.e. theory truly plays the role of a system of interpretation of data. There is therefore no contradiction in principle between the hermeneutic approach to cognition and the scientific imperative of functional modeling of cognitive mechanisms.

Second, affirming the hermeneutic nature of symbolic cognition does not mean accepting that there can be no communication among people or that cultures are incommensurable, on the pretext that “everything is a matter of interpretation”. It is understood that certain messages are produced with the intention of being interpreted in a certain way and not any other way. The thesis of the hermeneutic nature of cognition in no way precludes interlocutors from different cultures from sharing common interpretation procedures. This is precisely the case with mathematics and the exact sciences, whose statements and practices can circulate among different cultures. In music, dance and other art forms, the expressive force also seems to transcend the barriers of cultural codes. Finally, international trade, in spite of its risks and inequities, is further evidence of cross-cultural agreements that work.

Third, as I stated above, radical hermeneutics does not necessarily imply extreme relativistic indifference. It is clear that certain methods of interpretation, because of their practical consequences, are not viable in the long term, while others tend instead to benefit the communities or institutions that use them.

Now that possible misunderstandings have been dissipated, I can say that radical hermeneutics states simply that symbolic cognition (whether individual or social) involves processes of creative interpretation rather than representations of a pregiven reality. There is no categorization, evaluation, narration or theorizing that is absolutely true: all these operations interpret the given, i.e. they give it meaning. In addition, in human cognition the production of meaning is necessarily based on cultural apparatuses. It is indissociable from social memory, in which meaning emerges from the collective manipulation of symbolic systems (the classic examples of which are languages and writing systems) working on corpora of shared data.

I conceived the cognitive model of the Hypercortex so that it would be in keeping with the hermeneutic approach to cognition I have described. In this model, IEML plays the role of language for data interpretation, or metalanguage. The metalanguage is shared, but the statements (the acts of categorization) and the texts in IEML (the USLs) are unique and are the responsibility of their authors. Categorization is not all that is involved. It is also necessary to consider evaluation. In fact, as I pointed out above, the production of meaning does not occur without a certain intensity of emotion or force of intention. Could there be any distinctly human meaning without circulations of affects and desires infusing their energy into the games of symbolic structures? That is why the semantic machine at the core of the Hypercortex automatically transforms IEML texts (USLs) into circuits in which evaluations will produce semantic currents. These semantic flows represent the affective dimension of the production of meaning.

13.4. The hermeneutics of information

13.4.1. Data

The base of our pyramid of the complexity of memory is occupied by data. From the point of view of cognitive modeling in the Hypercortex, data are not defined by their content but by their addressing system. Data can therefore be rudimentary or elaborate, quantitative or textual, audio or video, light or massive: they will be considered data only because they are addressed by URLs. It should be noted that the URL is in no way a unique description of unique content. The same content can be found in many different URLs and the same URL can address a flow of data rather than a fixed content, as is commonly seen in blogs and newsfeeds of social media. The URL is opaque and simply provides an access route to the electronic container, but it says nothing about the content of this container.

13.4.2. Perception

Ideally, creative conversations should be able to collaboratively categorize, evaluate and filter the storehouse of data of the Web according to their own criteria. In addition, if data are organized in different ways, so that creative conversations are separated by walls, the potential usefulness of a global memory is not optimized. Let us recall the well-known silos created by the incompatible formats of the “clouds” controlled by the big companies of the Web or the “semantic silos” of ontologies. The problem is going from data (identified by the addresses of their electronic containers) to semantic information units, i.e. data freely categorized and evaluated but meeting the requirements of semantic interoperability. The function of perception is implemented by a mechanism that adds a metadatum (a semantic current in the semantic circuit corresponding to a USL) to a datum. When the datum “perceived” by the cognitive system of a creative conversation.

13.4.2.1. Categorization

In the model of hermeneutic memory, a datum is defined solely by its Web address (a URL). To obtain a semantic information unit, this datum must first be categorized, i.e. given a semantic address, a USL, with which the STAR-IEML linguistic engine automatically associates a semantic circuit. Assigning USLs to data can be done by any method imaginable, from the most spontaneous, artisanal and “manual” to the most regular, industrial and automated²⁴. There is no question here of any kind of regulation.

13.4.2.2. The semantic current

The strength of the link between a multimedia datum (formalized by as a URL) and the semantic address that categorizes it (formalized as a USL) is represented by a semantic current C²⁵. The semantic current is a symbolic energy insofar as it joins together the two parts (category and datum) of a symbolic unit. It is the energy that, in connecting the semantic sphere to the web of data, creates the information unit. The production of this current can itself be broken down into two components: the production of polarity and the production of intensity.

13.4.2.3. The production of polarity

After categorization (“What is it? What is it about?”), the second major function of assimilation of data in symbolic cognition is thus its evaluation, which can be broken down into the evaluation of quality (typically, good or bad) and of quantity (typically, little or a lot). The quality, or polarity, of the current (“What is it worth?”) is represented by an ordinal number²⁶. The function of evaluation of polarity corresponds roughly to the affective dimension of animal cognition: attraction or repulsion, pleasure or pain. Polarity can be determined by a price or by a vote, by manual or automatic procedures, etc. All games of evaluation are possible. The word polarity designates nothing absolute here. It is a degree on a scale based on certain evaluation criteria. The value of polarity can indicate truth (for example, on a scale from completely false to completely true), importance, suitability, danger, humor, beauty, effectiveness, etc., as the case may be. Polarity thus has an assignable meaning only if the method and criteria of evaluation used to determine it are explicit. Like the functions of categorization, the functions of evaluation of polarity should be absolutely free. They can meet a large number of distinct criteria, even a whole range of combinations of criteria.

13.4.2.4. The production of intensity

A clear distinction needs to be made between the qualitative evaluation of polarity and the quantitative evaluation of intensity. This is not always obvious because we tend to believe that everything that has a numerical form is quantitative. When a teacher gives a student a mark, for example, it does not really represent a quantity, although it is a number. It is simply a convenient way to place the student on a scale of quality or value in comparison to other students in the same class or in relation to a criterion of excellence established by the academic institution. It is clear, for example, that the question of whether a document is of high or low quality is distinct from that of whether the document contains little or a lot of text, measured in bytes (quantity). The intensity of semantic current measures quantity and is formalized by as a cardinal number. It may indicate the number of downloads or clicks, the volume of a data flow or frequency of use, but also, depending on the way a creative conversation creates its information units, a volume of production or consumption, a debit or a credit, etc. Like the evaluation of quality, the production of the intensity of semantic current should follow an explicit procedure.

13.4.2.5. The result of perception: the phenomenal information unit

The operation of perception combines two hermeneutic operations: categorization, which determines the semantic circuit associated with a datum, and production of semantic current in the circuit. The production of the current can itself be divided into the production of intensity and the production of polarity²⁷. The result of perception is a phenomenal information unit, or a phenomenal idea.

13.4.3. The semantic information unit

As shown in Figure 11.5, the semantic information unit is represented by a triad (URL, C, USL). First, the semantic information unit so defined represents a datum that is actually categorized, quantified and evaluated. This is precisely what we were looking for. Second, each creative conversation can adopt its own rules of perception. The functions of categorization and production of current on the data of the Web are entirely free, and are thus in keeping with our hermeneutic approach to cognition. Third, all semantic information units are interoperable, since they are expressed in interoperable terms. Indeed, (i) URLs are universal; (ii) USLs and numbers, also universal, are variables of calculable transformation groups; and, finally, (iii) USLs are translated automatically into semantic circuits that are readable in all languages. Our model of ideas – semantic information units – is thus in keeping with the epistemological requirements of full explication and calculability of the exact sciences.

13.5. The hermeneutics of knowledge

13.5.1. Thought

The data categorized and evaluated by creative conversations can be considered their perceptions or phenomenal ideas. Once these phenomenal ideas are produced, the problem of creative conversations is to put them to use in order to understand their environment and orient their actions. This is where the functions of thought come in.

Unlike what happens with the functions of perception, the input variables of the functions of thought are not data (URLs), but semantic information units (USL, C, URL), which are produced either by functions of perception or by functions of thought.

The output variables of the functions of thought are circuits of semantic current among information units: narratives, sequence of statements, activation of networks of ideas, simulations, etc. I call these circuits noumenal circuits, and the information units that make them up noumenal information units or noumenal ideas. Semantic circuits, which are graphs of USLs, should not be confused with noumenal circuits, which are graphs of semantic information units (USL, C, URL). Moreover, noumenal ideas have exactly the same composition (USL, C, URL) as phenomenal ideas; they are just produced by different functions.

The functions of thought interpret the information units that emerge from the perception of data or from other functions of thought. When I say that the functions of thought interpret the phenomenal ideas resulting from perception, I do not mean to imply that these ideas are neutral and are not themselves results of interpretative processes. Quite the contrary! Phenomenal ideas are indeed products of hermeneutic functions, and functions of thought in turn interpret these products. One of the roles of functions of thought is to situate phenomenal ideas in (supposed or imagined) patterns of emergence, transformation and disappearance of phenomena. In short, they place information in relationships.

Noumenal circuits (in the vocabulary of the IEML model) can be considered to formalize knowledge (in the vocabulary of information management). Unlike the rather vague descriptions of the DIKW theory, the term noumenal circuit here has a very precise technical meaning: it is a network of relationships among semantic information units, a network that is produced by an explicit function of thought. Once again, the essential point is that the USLs and the semantic current C that make up information units (USL, C, URL) are variables of transformation groups. Functions of thought can thus automatically generate oriented graphs of USLs and transformations of the semantic current in these graphs.

13.5.2. The semantic information unit as a tool for cognitive modeling

The above overview of hermeneutic functions has provided a general idea of the production, use and placement of information units in circuits by IEML collective interpretation games:

– These information units are produced from data (URLs) by functions of perception, and from other information units (USL, C, URL) by functions of thought.

– They are used (as input variables) and interconnected in noumenal circuits by functions of thought.

I will now translate the formal model of the semantic information unit into the cognitive registers of ideation, enunciation and memory.

13.5.2.1. The information unit as an idea

The semantic information units (URL, C, USL) model the phenomenal and noumenal ideas of hypercortical cognition. A comparison of Figure 9.2 and Figure 11.5 shows that the formal structure of the semantic information unit is also the structure of an idea: (T) multimedia data (URLs) represent sensory data, or percepts; (S) USLs represent concepts that categorize data; and (B) the semantic current C represents the value (attraction or repulsion) of the categorized percept, namely the affect.

It should be noted that this objective unit of hypercortical cognition is not subject to any limitation of scale. An abundant source of data (represented by a URL) categorized as a semantic circuit (since the USL represents a circuit) as broad and complex as desired, in which the distribution of the current is transformed according to the variation of the data, can be conceived as a single variable idea. The data flow can come from a traditional site, an object, a sensor or a news originator. Here, the semantic information unit indicates the meaning and relevance of the data flow identified by its URL. This approach is particularly suitable for the “web of flows” or “real-time web”.

13.5.2.2. The information unit as an utterance

The semantic information unit can also be considered the model of a referenced utterance. According to this perspective, the discourse or utterance is supplied by the USL, the reference by the URL, and the pragmatic force of the utterance by the semantic current C that links the URL and the USL through a function of explicit evaluation. In this approach, hermeneutic functions can be considered functions of enunciation, since they produce referenced utterances driven by a pragmatic force. When an information unit is dated and signed, it becomes a completely explicit “unit of enunciation”. The player, user, person, community or creative conversation that takes responsibility for the enunciation can be considered the author of the utterance.

13.5.2.3. The information unit as a meme

Finally, semantic information units can also be considered units of memory, or memes. It should be noted, however, that these memes are much more elaborate than those of Dawkins’s memetics, which is based on a biological model that is inadequate for the complexity of cultural processes. The hermeneutic memory of the Hypercortex, seen as a holistic cognitive faculty, simultaneously represents the potential for both memory and forgetting. Each creative conversation decides what it retains and what it disregards, according to its own criteria of perception and thought.

In the foreground of its cognitive mirror, the creative conversation conducts a dance of urgent, important information: its “here and now”.

In the middle ground, it displays the familiar information units and knowledge it has to be able to recall quickly in order to understand its present.

In the background is the rest of the information, structured according to its own hermeneutic functions, which must be accessible in one way or another because it may be useful or interesting.

Finally, in the shadowy depths of its unconscious is the opaque ocean of unperceived, unthought-of data or data interpreted according to criteria too different from its own to be useful to it.

The creative conversation thus projects its own cognitive map on the semantic sphere, a map arranged in concentric strata, from well-lit foregrounds to shadowy backgrounds and, further, to the gradual darkness of forgetting. For each creative conversation, the landscapes of memory and forgetting are different, the gradations of reflexive consciousness and total unconsciousness are folded according to other folds. These conversations can dialogue, learn from each other and even merge, intersect or differentiate their memories at will, however, because they share the same semantic sphere, the same transformation group. Woven of fractal circuits among information units, subjected to intense currents and emotional storms, the whole semantic sphere turns and reorganizes itself around creative conversations, obeying their collective interpretation games.

13.5.3. The noumenal circuit as a tool for cognitive modeling

Having examined semantic information units, I would now like to consider the use that may be made of the noumenal circuits of hermeneutic memory for cognitive modeling. I cannot prejudge all that the collective intelligence of creative conversations will invent in the area of functions of thought, so I will just mention a few possible uses and suggest some directions for development.

13.5.3.1. The noumenal circuit as a theory

A piece of knowledge (a noumenal circuit) functions as the context that explicates the information units it links. Since this context has been constructed by a freely chosen function, there is obviously nothing “natural” about it; it is an interpretation. From an epistemological perspective, the function of thought can be compared to a theory that produces knowledge based on phenomenal ideas. Thus not only can a single multimedia datum enter into the fabrication of many different phenomena, but the same phenomenon can also be explicated by many different theories, i.e. by many ways of relating phenomena to each other. Theory simulates relationships among phenomena (among phenomenal ideas). That is why functions of thought can be considered tools of cognitive simulation, capable of producing useful knowledge for creative conversations.

13.5.3.2. The noumenal circuit as narrative

Cognitive psychology long ago taught us that one of the best ways in which to retain information is to organize it in narratives. While semantic information units are produced by acts of enunciation, noumenal circuits (i.e. knowledge in ordinary terms) become circuits of enunciations, which can be seen as complex narratives. According to this approach, a function of thought can be processed as a narrative function that arranges relationships among acts of enunciation. Once again, the authors of these narrative functions are free to place – i.e. interpret – the same enunciation in narratives that may be completely different.

13.5.3.3. Cognitive simulation

Perception²⁸ of data can be compared to the “sensory excitation” of a set of USLs by a semantic current. However, this excitation of USLs represents only the initial by the USLs that act as semantic sensors can be processed by all kinds of algorithms. Noumenal circuits can channel the transformations of the semantic current received by the sensors. The sensory input of the Hypercortex can then be transferred and processed along the noumenal circuits to which the receptor USLs are connected. The current can be amplified, blocked, summoned or freed according to thresholds. It can propagate through resonance and simulate cognitive processes that can be as complex as you wish. Massively parallel distributed calculations (of the “neural network” type) can be constructed by supplying USLs assembled in circuits with automata for processing semantic current²⁹. All types of functions of thought are imaginable, not only those based on neural networks: fluid dynamics, heat propagation, genetic algorithms and artificial life programs, functions simulating emergent cognition in certain animal societies (swarm intelligence), classic economic games, logic rules of all kinds, not to mention functions drawing on the arts, humanities and social sciences to describe, invent or simulate original forms of collective cognition and dynamics of actor networks as closely as possible³⁰.

Finally, the noumenal calculations of the Hypercortex, as they are freely determined by creative conversations, can result in output information units, which can be considered “effecter excitations”: robot control, production of multimedia data for users, syntheses, predictions, etc.

In short, the functions of thought on semantic information units model the ecosystems of ideas discussed in Chapter 6.

13.5.4. Hierarchy of the functions of symbolic cognition

I will now review the different cognitive functions that can be formalized and automated by the IEML Hypercortex.

13.5.4.1. Semantic functions

Textual functions produce and transform USLs (texts in IEML).

Linguistic functions transform USLs into semantic circuits that are readable in natural languages, and vice versa. They can also translate a semantic circuit that is readable in natural language x into a semantic circuit that is readable in natural language y.

Conceptual functions produce, transform and measure semantic circuits readable in natural languages.

13.5.4.2. Hermeneutic functions

Functions of perception create semantic information units (USL, C, URL) from data (URLs).

Functions of thought create noumenal circuits among semantic information units, using semantic information units.

The automation of semantic functions is the basis for the automation of hermeneutic functions, those that create semantic information units (URL, C, USL) and those that produce knowledge (noumenal circuits). There can be no Hypercortex, no universal digital memory serving collective intelligence, without automatic and conventional transformation between USLs and semantic circuits interpreted in natural languages. Indeed, the availability of USLs that have meaning is an essential condition for the automatic creation of semantic information units and their circuits. The existence of the IEML semantic machine is therefore the basis of the possibility of implementing a hermeneutic memory. USLs, semantic circuits, semantic information units and noumenal circuits are automatable constructions. While automatable, however, they are no less free, transparent and are hypothetical. As such, like all hypotheses, they can be deconstructed.

13.6. Wisdom

Nothing will ever rule out the categorization of data or the placement of information units in context “by hand”. All my effort to describe the cognitive model of the Hypercortex is to show that these operations are automatable and that this automation can augment the collective intelligence of online creative conversations. The programming of the functions of perception or thought, however, is not itself automatable. It depends on the free decisions of the communities concerned and, more generally, on a practical wisdom that Aristotle called phronesis, which is often translated as prudence³¹. By explicating – and thus augmenting – the processes of collective cognition, this wisdom pursues a goal that is concrete rather than contemplative: it is in keeping with the actual needs of a community. Its good (or its striving toward improvement) is to be found in a middle course between excesses and deficiencies. It confronts problems of degree and balance: how can we measure, evaluate, categorize or generalize without putting too much importance on minor data, without disregarding “weak signals” and without missing the essential? How can we place information in context, but without getting submerged in generalities? This wisdom of the middle way does not involve blindly following the majority or submitting to a statistical average. On the contrary, it demands firmness, courage and independent judgment, not to mention the capacity to criticize your own decisions in light of their results. While such practical wisdom can and should produce deterministic rules – those of the collective interpretation games – its private operation cannot itself be based on deterministic rules. It will, at most, call upon heuristics, methods of stimulating invention and openness to situations. If knowledge is the organization of memory, wisdom is the organizer. At the top of the DIKW hierarchy, wisdom must confront the problem of the arrangement of a memory. How can functions of perception and thought (which will determine methods of filtering and searching) best suited to the needs and desires of a creative conversation be created? How can we construct information units from data, and knowledge from information? What are the organizing narratives that give meaning to ideas?

Information is an interpretation of data, and knowledge is an interpretation of information. That is why the wisdom that governs the cognitive operations of memory cannot come out of any absolute precept or supposedly objective science. It is a hermeneutic wisdom, and thus eminently free and open, which proposes conventions to particular communities rather than offering transcendent, universal truths. The historical experience of humanity shows that hermeneutic wisdom is constructed patiently along paths interwoven with traditions of interpretations of secular corpora³². This wisdom requires a long memory, because it is responsible for thinking about the long-term effects of the way our creative conversations create their information units and extract knowledge from them. The ultimate purpose of hermeneutic wisdom is to improve the cognitive functions of communities that are organized around a shared memory, and to do so within a sustainable horizon of learning and discovery.

13.7. Collective interpretation games

Before closing this chapter on the hermeneutic memory of the Hypercortex, I would like show how collective interpretation games can become social tools. The design of these games can indeed be considered an art or, if you prefer, a wisdom that serves the lasting augmentation of the cognitive power of communities of players.

Collective interpretation games cause interaction between automatic functions and human perceptions, actions and reactions in self-organizing loops. The automatic functions can be divided into two major classes: hermeneutic functions, which we have already studied in this chapter; and multimedia navigation functions, which make it possible to explore the universe of information organized by the Hypercortex in an interactive, polysensory way. The game is based on the process of continuous feedback from the players, who produce new data and new metadata and progressively develop their hermeneutic functions and navigation systems. A collective interpretation game should be considered the hypercortical avatar of a creative conversation.

13.7.1. Reading/writing

We can imagine iconic or musical translations of USLs in addition to their translation into natural languages. Instead of starting from writing a text using a keyboard, the construction of semantic circuits could be controlled by manipulations of symbolic objects in augmented reality, using digital capture of movements. Reading could also be augmented by sensory-motor exploration of hypertext circuits whose nodes were represented by figurative objects (instead of texts in IEML or natural languages).

13.7.2. Exploration

To access the cognitive processes simulated in the Hypercortex, all multimedia interfaces are possible. The various phases of automated hermeneutic processing of data, as well as their results, can be communicated to the players by means of interactive and immersive representations using virtual or augmented reality. Drawing on certain video games, it would be possible to project processes modeled in the semantic sphere into an immersive 3D space. Semantic browsers should facilitate collaborative exploration of the semantic sphere and make players aware of regularities and singularities of hypercortical processes.

13.7.3. Feedback

Once they are aware of the cognitive processes that are modeled by the Hypercortex and represent the state of the collective interpretation game in which they are participating, users are capable of producing data and/or IEML metadata in order to respond to the situation. A collective interpretation game therefore lives along a communication loop. This loop begins with the distributed production of data. It continues with the production (also distributed) of metadata: categorization, evaluation and quantitative measurement of the data. The metadata are projected in the semantic sphere, which causes associative or contextual automatic processing: semantic currents flow through the noumenal circuits of the semantic sphere and transform the moving landscape of meaning of the Hypercortex. The results of the processing are then sent back to the players on their mobile devices or through an immersive multimedia environment in augmented reality. The signals from the semantic sphere synthesize the dynamics of shared memory resulting from the actions of all the players. On the basis of their interpretations of these signals, the players can then produce data – symbolic physical actions, movements or expressions – that will steer or stabilize the collective interpretation games in the direction desired.

In an environment enriched with ubiquitous computing and robots, players can at any time consult the current representations of their favorite games. They use these games to coordinate their activities, improve their information searching and filtering, optimize their learning activities to meet their needs, organize and synthesize huge masses of information quickly, manage their social networks, navigate in urban landscapes, orient themselves in geographic space and, finally, make decisions that will be reflected in shifts of data flows or the evolution of hermeneutic functions. In performing all these tasks, players benefit from information and data from the interaction of all the other players, whatever their languages or home institutions, and do so transparently, regardless of the material platforms (computers, mobile devices or cloud services) they use.

13.7.4. Coordination of the games

The collective interpretation games based on the IEML semantic sphere function as mechanisms for the internal organization of creative conversations thanks to a ubiquitous distributed memory with symbolic processing capacities using the full computational power of the digital medium. A collective interpretation game thus serves as an instrument of both observation and steering of creative conversations: a virtual vessel exploring the semantic sphere in its own way. We need only look out of the porthole to see a swarm of other vessels with which its explorations can be coordinated. The different games can form relationships since they share the same semantic sphere. Beyond knowledge management and cognitive augmentation, the machinery of the Hypercortex can and should be used for all kinds of scientific modeling and simulation of cognitive processes, at the level of both individual cognition and that of communities and cultures.

The information economy (which gives its name to the IEML language) coordinates all the collective interpretation games that drive the Hypercortex. All the games are in competitive cooperation in a general economy of information in which data, semantic information units (categorized and evaluated data), semantic circuits, noumenal circuits and functional modules of cognitive models are exchanged. It is this semantic information economy – an economy whose hermeneutic functions are transparent and freely chosen – that now makes the scientific exploration of the human cognitive ecosystem as it is expressed in the digital medium possible.

---

1 As we will see, this chapter deals with the same subject as Chapter 6 and contributes to solving the problems raised in Chapters 4 and 5.

2 See The Savage Mind [LÉV 1966].

3 See the excellent book by Geoffrey Bowker, Memory Practices in the Sciences [BOW 2005].

4 See the classic book by Lev Manovich, The Language of New Media [MAN 2001].

5 On all these topics, see Nova Spivack’s blog: http://www.novaspivack.com.

6 See Chapter 4 on creative conversation.

7 See the technical studies by Smith, Tagging: People-powered Metadata for the Social Web [SMI 2007], and Dichev et al., “A study on community formation in collaborative tagging systems” [DIC 2008]. Also worth consulting is Michèle Drechsler’s thesis, Le Socialbookmarking dans l’Éducation [DRE 2009].

8 OWL is the acronym for Ontology Web Language. See Feigenbaum et al., “The semantic web in action”, and Handler and Allemang, Semantic Web for the Working Ontologist [FEI 2007, HEN 2008].

9 [ELI 1934].

10 See, for example, [ACK 1989].

11 The four layers of addressing of the digital medium (see Figure 12.3) must not be confused with the four layers of complexity of memory, which I am discussing here (see Figure 13.2).

12 See Georges Gusdorf, Les Origines de l’Herméneutique [GUS 1988].

13 See Philosophical Investigations, [WIT 1958]. Wittgenstein does not use the word text, but his approach to language games and their “grammars” could be translated into a general theory of textuality.

14 Michel Foucault, The Archaeology of Knowledge [FOU 1972]. In this book, Foucault develops the concept of discursive formation and extends the concept of episteme. Judgments of truth and interpretive statements are reduced to the sociohistorical conditions in which enunciating subjectivites (I would say: cognitive functions) are constructed. Although Foucault emphasizes the multiplicities, ruptures and differences that act on discursive formations, he is clearly drawing on Nietzschean perspectivism (therefore hermeneutics) and structuralism.

15 See Of Grammatology [DER 1976], which puts forward an abstract view of writing or text and sees it as prior to – and independent of – any system of characters, letters or notation of speech.

16 See Jean-François Lyotard, The Postmodern Condition: A Report on Knowledge [LYO 1984] and The Differend: Phrases in Dispute [LYO 1988].

17 Hans-Georg Gadamer, Truth and Method [GAD 1988].

18 Paul Ricoeur, The Conflict of Interpretations: Essays in Hermeneutics I and From Text to Action: Essays in Hermeneutics II [RIC 1974, RIC 1991].

19 See the remarkable Etica dell’Interpretazione [VAT 1989], in which Vattimo clearly distinguishes radical hermeneutics, which originates in a meditation on language, from plain and simple cultural relativism.

20 Jean Grondin, L’Universalité de l’Herméneutique [GRO 1993] and L’Herméneutique [GRO 2006].

21 It is useful here to remember that as a very young man, Nietzsche was a teacher of philology and that he had been trained in the interpretation of ancient texts [NIE 1900].

22 I am speaking of prudence in Aristotle’s sense, i.e. as practical wisdom that comes from moral strength, a courageous prudence that has nothing to do with fear or timidity.

23 Cornelius Castoriadis, The Imaginary Institution of Society [CAS 1998].

24 All kinds of methods of statistical analysis of data or software for automatic processing of natural language can be used for categorization in IEML. See, for example, Yair Neuman and Ophir Nave’s interesting method for extracting meaning from documents based on the analysis of metaphors [NEU 2009].

25 On the concept of semantic current and its justification from the point of view of the human sciences, see section 6.3.

26 An ordinal number designates a “rank” or numerical order. It gives the position of an item in an ordered series, while a cardinal number is used to define the size of a set. On the importance of ordinal numbers to express priorities or preferences, see [SLO 2009].

27 See Figure 13.3.

28 Perception, i.e. categorization plus the measurement of value and intensity.

29 The forerunner of this type of calculation was Warren McCulloch [LÉV 1986a, MAC 1965]. Subsequently, von Foerster and his team at the Biological Computer Laboratory [FOE 1981, LÉV 1986b] developed McCulloch’s ideas. Massively parallel calculation based on neural networks has become a well-established sub-discipline of artificial intelligence; see McClelland and Rumelhart, Parallel Distributed Processing: Explorations in the Microstructure of Cognition [MAC 1986].

30 On the dynamics of actor networks and the modeling of technical, cultural and social phenomena using graphs, see my comments on Figure 5.1.

31 See Aristotle, Nicomachean Ethics, in particular Book VI. See also Pierre Aubenque, La Prudence chez Aristote [AUB 1963]. Practical prudence (phronesis) is contrasted with theoretical wisdom (sophia).

32 Corpora is the Latin plural of corpus.