De l'espirit entendu a l'intelligence collective

Harry Halpin - Researcher at IRI (Institut d'Recherche et Innovation) - MARIE CURIE Project PHILOWEB exchange with World Wide Web Consortium/MIT CSAIL, harry@w3.org

Textorized Parthenon

De l'espirit entendu a l'intelligence collective

The Cartesian Presupposition

Although much ado has been made in the cognitive sciences over Cartesian dualism, what has until recently been unquestioned in Descartes and both his materialist (Locke, Hobbes) and idealist contemporaries (Berkeley, Kant) is their ontology.

All strands of Enlightenment thought start with the I, the individual subject whose brain (and thus, mind) is the "engine of reason and seat of the soul". Regardless of the details of how cognition is implemented, philosophy since the Enlightenment, in particular analytic philosophy, in has often assumed that the (a priori) unit of analysis is the human individual, which interacts with the world via their sensory-motor capabilities. In the Kantian tradition of classical cognitive science, these sensor-motor capabilities create representations of the external world as internal representations. In current embodied approaches to cognition, the external world is dealt with directly via non-representational neural processes interact with the world without the need for detailed internal representations. Regardless, we can assure ourselves that I think therefore I am.. Materialism has won the day, yet strangely co-inciding with the advent of ubiquitous digital technologies, all is not as stable as it seems.

The Digital Turn

The concept of the individual is breaking down in the 21st century. This breakdown is in part due to both the massive proliferation of external representational media upon which we increasingly depend. While once confined to perceptual-specific media (pictures and photography, radio, television), due to the invention of digital computers the amount of media is growing at an astounding rate, with the growth of digital media outpacing all previously recorded analogue media. While the question of the existence of internal neural representations is still ultimately up for debate, there is no denying our increased interaction with an increasingly digitized and symbolic external environment, where much of the "reasoning" is no longer confined "inside the head" but is done on our behalf by various automatized algorithms. The View from Nowhere

Not only has the amount of digital information grown without bounds, but there has been a massive decrease in the latency necessary to access digital information due to the now ubiquitous spread of high-speed Internet access. Due to mobile smartphones, the current generation of "digital natives" has always been effectively online, and not only are the psychological and neurological consequences of this unknown, but so are the ontological consequences. While once one had to go to a library to physically retrieve a book from a shelf or even put a compact disc into a player in order to listen to music, today it appears as if all of the digital and digitized information produced by humanity can be within the grasp of Internet-connected human within seconds if they are familiar with the retrieval engines such as Google. Today, you no longer even need to search: Thanks to services ranging from Twitter, customized algorithms can instantly and even without interaction bring digital content to that you might want.

The Extended Mind Hypothesis

The question is then: Does cognition - the mind itself - extend into the environment? Two analytic philosophers of the mind, Clark and Chalmers present the example of Otto, a person with short-term memory loss who has to navigate their way to the Museum of Modern Art (MOMA) in New York without the ability to remember an address. Instead of relying on his internal memory, Otto has to rely on his notebook. Given that the notebook performs the same functional role as neural memory, should the notebook be given some of the cognitive credit despite being outside the skin? (Clark and Chalmers, 1998)

The View from Nowhere

The Extended Mind Hypothesis states that cognitive processes, under certain conditions, are constituted by the environment. Thus, an extended mind is a new integrated cognitive whole or "cognitive system" that requires a mutually-reciprocal relationship between the environment and the organism as regards some cognitive function such as memory or attention.

The Extended Mind Hypothesis is much more radical than the Embodied Mind Hypothesis, which in contrast states that cognitive processes depend very heavily in hitherto unexpected ways, on organismically external props and on the structure of the external environment which cognition takes place. The key is that an extended mind is constituted by some aspects of the world "outside the skin," while an embodied and embedded mind merely is deeply causally intertwined with the environment. In other words, if Otto loses his map, he may lose some of his mind! (Rupert, 2004)

Clark also provides what he considers four principles to distinguish cognitive extension from mere causal coupling: portable, robust, dovetailed (co-adapted), augmentation (extends capabilities)

The Web-Extended Mind

Imagine the case where Otto is not only using a map to go to the Museum of Modern Art, but is being guided by a Web-enabled smartphone that is updated in real-time. If there is a particular subway closure that may prevent him from getting there, real-time notifications as done by either crowd-sourced human beings can warn him and guide his behavior. Is then the Web part of Otto's extended mind? Does losing your smartphone mean that you may lose some of your mind? Let us inspect the Web in light of Clark's criteria:

The View from Nowhere
  1. Portable With the advent of smartphones and (soon) devices like Google Glasses coming, access to the Web is now portable.
  2. Robust With increasing connection speeds and coverage for Web access, the connection should always work.
  3. Dovetailed Personalized algorithms that learn from behavior such as recommender systems and search engines are co-adapted to their users.
  4. Augmented The speed at which one can retrieve information and communicate extends current capabilities (such as pre-Web capabilities of linguistic reading and speech) to such an extent it counts as a new capability (Berners-Lee and Halpin, 2012)

Indeed, not always but under some circumstances the Web itself can be considered to be part of the mind. Just visiting and reading a web-site from a desktop computer is not cognitively extended. An always-accessible smartphone that based on your geolocation and current set of tasks delivers you relevant information "just in time" is part of your cognitively extended machinery.

Cognitive Integration:

What prevents any part of the world with a merely causal relationship to the human body from having cognitive capacity is that Clark counts as cognitive only those parts of the world that are constitutive of a new integrated cognitive whole - a system with cognitive capabilities. This requires a mutually-reciprocal relationship between the environment and the organism as regards some function such as memory.

Example: In experiments, it has been found that students who were given difficult problems and allowed to gesture with their hands (either spontaneously or encouraged explicitly) performed better than those who did not gesture. (McNeill, 2005)

Casual-cognitive/coupling fallacy: Question: Why did the pencil think that 2+2=4? Clark's Answer: Because it was coupled to the mathematician (Adams and Aizawa, 2001).

Response from Clark: Why did the V4 neuron 'think' there was a spiral pattern in the stimulus? Because it was coupled to the rest of the monkey!

Controller Objection: Even if external elements sometimes participate in process of control and choice (your software agent might chose some stocks and shares, and so on) still it is always the biological brain that has the final say.

Response from Clark: But do my frontal lobes have the real final say? Then doesn't my self shrink?

Cognitive Self-Stimulation: Cognitive self-stimulation occurs when a system is causally responsible for producing structures and events which are then recycled as inputs to the very cognitive system itself and thus sustains sophisticated (brain-body-environment) loops of exploitation and mutual co-ordination that can help solve particular problems. The boundary between an "input" (perception) to a system and the "output" (action) of a system melts and the interface becomes part of the system.

Egocentric Speech

Egocentric speech is the the ability for children to speak to themselves when they are accomplishing some task. This phenomenon seems to begin at seven or eight years of age and then disappears afterwards. Is egocentric speech an example of cognitive self-stimulation?

The View from Nowhere Egocentric speech is "externalized" when children encounter problems they seem to need help solving. In the presence of solving difficult tasks, the amount of egocentric speech in children increases. This can even be seen in adults when they "talk through a problem" - even to themselves. Why does this externalization help? It can be considered that egocentric speech is helping direct the child's actions, in the same way that a parent's voice directs a child.

Piaget assumed such speech was purely self-directed and individual: To put it quite simply, we may say that the adult thinks socially, even when he is alone, and that the child under seven thinks egocentrically, even when in the society of others. (Piaget, 1959).

Vygotsky reinterprets these findings: Externalization of speech is not only necessary to direct the child (perhaps due to the lack of language being internalized) and thus part of a cognitively integrated child. Strangely enough, Vygotsky showed that cognitive self-stimulation via egocentric speech is actually social: In experiments with deaf-mute children or with children speaking a foreign language...the coefficient of egocentric speech dropped to zero in the majority of the cases and the rest to one eighth. (Vygotsky, 1986).

The Social Becomes the Cognitive

Egocentric cognitive self-stimulation is social: Children who are participants in the collective monologue do believe that they communicate with each other. They believe that their thoughts, even those that are poorly expressed, or unarticulated, belong to all participants.(Vygotsky, 1986)

What happens is between the ages of five and eight egocentric speech becomes internalized as inner speech and thus transforms the capacities of thought, so that thought can possess various linguistic characteristics such as abstraction and systematicity. Speech is ultimately external before convergence with internal thought. Egocentric speech reflects the imitation of the social guidance of parents, even if other children can not understand the speech.

Vygotsky's theory is that functions are first external and social before becoming part of the mind: Any higher mental function necessarily goes through an external stage in its development because it is initially a social function. This is the centre of the whole problem of internal and external behvior...when we speak of a process, "external" mean "social." Any higher mental function was external because it was social at some point before becoming an internal, truly mental function. It was first a social relation between two people. (Vygotsky, 1978)

To go beyond Clark, a fully-developed individual is the product of a process of collective individuation (Stiegler) that requires continuous feedback learning-loops and dynamic soft assemblages between a biological body, technology, and other people.

The Social Web

There is the strange paradox that while most traditional philosophers find the Extended Mind Hypothesis disturbing, computer scientists find the idea of having their mind extended into the Internet to be common-sense. Is this a delusion or a new ontology?

Tim Berners-Lee provides an excellent example: When Tim is thinking through an idea, he finds it useful to phrase the concept as best he can in text and put the idea into an Internet Relay Chat channel, where others can discuss and comment on the idea, and so collectively improve and eventually code the idea. The View from Nowhere

The Web seems to be reversing our internalization of thought, with currently more and more thoughts being put on the Web for all to see and comment upon (and otherwise "mashup" and transform). Even if our thoughts are not understood, we nonetheless externalize them on the Web: "tweeting" on Twitter and "status updates" on Facebook can under many circumstances be considered digital variations of egocentric speech.

Yet does this Web-based communication lead to collective individuation? Perhaps not: When users are considered as social atoms superimposed onto a technological network, the spontaneity and innovation within their possible collective intelligence is deformed by the control of the networks, driven as it is by intensive marketing and consumerism aimed at individuals rather than the development of the potential of the group. (Hui and Halpin, forthcoming).

Varieties of Distributed Cognition

Edward Hutchins, an anthropologist studying human behavior in complex situations such as piloting a ship for the US Navy, observed that groups may have cognitive properties that differ from those individuals who constitute the group (Hutchins, 1995). The View from Nowhere Following Vygotsky, Hutchins holds that cognition is a cultural process. Each culture has over millenia developed solutions to frequently encountered problems are crystallized and saved in the material and conceptual tools of the trade and in the social organization of work. (Hutchins, 1995). Our integrated cognitive selves include not the immediate environment (active externalism) but also the causal history of traditional externalism insofar as our cognition depends on our ancestors transmitting their solutions to us.

While Clark extends the individual and Vygotsky has as the end-point of development the individual, Hutchins believes the combination of multiple individuals and their tools provides grounds for distributed cognition: When the nature of the problem is seen as a coordination amongst person and devices, much of the organization of behavior is removed from the performer and given over the the structure of the object or system with which one is co-ordinating. (Hutchins, 1995)

The Web, as a universal space of information (Berners-Lee), allows for the first time in human history the possibility of real-time global co-ordination via "sky-writing" (Harnad) to the entire world at once. This is possible as digital media provide a kind of media that supports, for many tasks, the sort of externalization of function better than others (Hutchins, 1995). It is no accident the Web is digital, as the digital is nothing if not robust, providing flawless copying and perfect reliability in a flawed and imperfect world (Haugeland, 1981).

Engineering Collective Intelligence

Collective intelligence builds on humans connected to each other (via technology) with ever decreasing latency. Yet there are dangers in building a cognitively integrated collective intelligence that is "connected" all the time. Hutchins began simulating distributed cognition via constraint satisfaction networks (where each node was a human and a link some kind of communicative connection) and discovered that more communication is not always in principle better than less. Under some conditions, increasing the richness of communication may result in undesirable properties at the group level (Hutchins, 1995).

Hutchins found that for highly connected groups with high bandwidth communication, the network moves towards the interpretation it had moved to in the absence of communication, but now it moves so more quickly...having arrived at that interpretation, they remain there, absolutely unmoved by any amount of evidence from the environment. At high levels of persuasiveness, this system thus manifests a much more extreme form of confirmation bias than any individual alone. (Hutchins, 1995).

So there is an argument in favor of not being plugged in all the time in order to allow evidence from the world to have sufficient diversity of interpretation of evidence. This is done by breaking up continuous high bandwidth communication...implemented in social organization, in the interaction of an individual with an external artifact, or through the use of internal mediating structure (Hutchins, 1995).

Collective intelligence is therefore not a singular opposition to the individual, but a landscape of distributed cognition. Extended minds that encompass more and more of humanity and their technology could exhibit what Stiegler called the toxicity of synchronicity. What is necessary is to understand the heterogeneous advantages and disadvantages of the varieties of distributed cognition; to learn what kinds of organizational forms, technologies, and social relationship can best create the kinds of collective intelligence necessary to tackle particular problems.

What is Collective Intelligence?

The term collective in collective intelligence is an empty place-holder, a mark of scientific and philosophical shame. A superficial understanding of collective intelligence simply posits some kind of aggregate in contrast with the individual: the individual versus their larger world, the individual against their crowd, the individual given the totality of existence. We only dimly grasp what the outlines of that which ontologically is to come after the individual.

The View from Nowhere

Clark provides one answer: a mind can, and always has been, extended into the world and thus technology. There is no individual mind per se, but a process of individuation that consists of a continually modifying soft assemblage of bio-technological components. Vygotsky points out that the process of individuation is always engulfed first and foremost in the social, and so what appears to be the cognitive that Clark "extends" outside the skin originates in the history of the social. Finally, Hutchins provides a yet another insight: that that groups and whole societies can exhibit genuinely new cognitive capabilities that are embedded in the larger networks of social relationships and technology.

What we are missing is the ability to "cut" the world at its joints, and so define what ontological category properly subsumes the place of the individual in the era of the Web? The clues to the answer may be found in work by Chilean philosophical biologists Maturana and Varela in their attempt to define a similarly difficult term - "life" itself - in terms of autopoiesis: "living organization is a circular organization which secures the production or maintenance of the components that specify it in such a manner that the product of their functioning is the very same organization that produces them" (Maturana and Varela, 1973).

The Open World

The problem with Maturana's notion of autopoiesis is the return of an arbitrary biological boundary: the skin of a human. Yet there is nothing about the definition of autopoiesis that binds it to biological individuals. An autopoietic system "through their interactions and transformations continuously regenerate and realize the network of processes (relations) that produced them", there is no reason a priori why these interactions have to be biological but can not include other humans and technology (Maturana and Varela, 1973).

The second part of this definition of autopoiesis does pose a real problem for advocates of collective intelligence, for it holds that the components that “constitute it (the machine) as a concrete unity in the space which they (the components) exist by specifying the topological domain of its realization as a network (Maturana and Varela, 1973).

Yet these membranes or boundaries also allow interaction with the outside environment! However, frogs eat gadflies and humans use the Web to get directions to Centre Pompidou. Thus, we find that examples like cognitive self-stimulation may open up autopoietic systems to not be eternal and unchanging biological species-beings, but open bio-technological assemblages.

Structural coupling defines an individual organism's interaction with the environment, a "history of recurrent interactions leading to the structural congruence between two (or more) systems" (Maturana and Varela, 1937).

Instead of two closed systems perturbing each other for their mutual autopoiesis, it is easy enough to change perspective to see them as one system maintaining a co-evolved autopoiesis that leads to, at points, cognitively integrated wholes that reproduce the very social relationships and and biotechnological conditions that exhibit the augmented capabilities of collective intelligence.

Defining Collective Intelligence

If the individual can be defined via autopoiesis, and to maintain its autopoiesis the individual can increasingly incorporate environmental components, then the individual is no longer a static, closed system, but an open and dynamic system capable of assimilating and decoupling from various components as it goes in and out of autopoiesis - and this includes digital representations. As the Web is one of the few technologies that satisfies the criteria of Clark's Extended Mind (portable, robust, co-adaptive, augmented) the Web is a natural medium for collective intelligence.

Then we can define collective intelligence as a form of autopoiesis that goes beyond the "skin" of a single individual as to create a self-reproducing and modifying unified intelligence that can include multiple individuals and aspects of their (possibly digital) environment. and so the explicitly open pharmacological approach of Stiegler is a crucial extension to the study of systems.

Collective intelligence requires that an individual mature to reproduce not only biologically, but their larger techno-social cognitive niche. This successful reproduction (which requires both connecting and disconnecting to others) is the key of long circuits necessary for Stiegler's collective individuation. Rather than a stable reproduction of either the individual human (Clark's biological skin-bag) or some pre-existing society (the stable society), the process of techno-social reproduction will always include a process of unstable innovation. This should be obvious, as the world itself undergoes change and so confronts any being with new problems that threatens their very ability to reproduce their world.

After three thousand years of explosion, by means of fragmentary and mechanical technologies, the Western world is imploding. During the mechanical ages we had extended our bodies in space. Today, after more than a century of electric technology, we have extended our central nervous system itself in a global embrace, abolishing both space and time as far as our planet is concerned. Rapidly, we approach the final phase of the extensions of man - the technological simulation of consciousness, when the creative process of knowing will be collectively and corporately extended to the whole of human society, much as we have already extended our senses and our nerves by the various media. (McLuhan, 1964)

Collective Individuation

The View from Nowhere

With the advent of the financial crisis and catastrophic climate change, never has the very reproduction of our world been so threatened, and yet never has it been greeted with such indifference. This reason is that the problems facing us seem literally impossible to even imagine: the magnitude of the task for coming generations dwarfs the cognitive grasp of any single individual and our pre-Web social institutions.

Intelligence has always been collective, yet the advent of the Web, a universal digital medium that contains within it the latent possibility of connecting all of humanity and the world as and when necessary, arrives just in time to disrupt our previously stable Enlightenment ontology of individual. However, a new scientifically-grounded ontology of collective intelligence is just one task amongst many for digital studies, for today we lack a thorough understanding of the transformation of humanity at the hands of our digital tertiary retentions, and such an understanding goes far beyond cognitive science.

Currently our technologies, from social networking sites like Facebook to search engines like Google, are inadequate to the scope of crisis, founded as they are on outmoded ontologies of the socially atomized individual and disembodied information. A new philosophically-informed engineering must be arise that can create the kinds of cognitive niches that foster collective individuation and harness collective intelligence to solve problems on a global scale. Even more importantly, our current ethics, based on the notion of the self-interested individual, is clearly is at fault for the current crisis; a new ethics must be re-invented that does justice to the the dense "intertwingling" between our sense of self, our technology, and our social relationships. It is not enough to define collective intelligence, or to learn how to (re)produce it, but we must also care for our intelligence and our world.

References