On the Organizational Roots of Bio-cognition

Abstract

The theme of this book is the place of organization in the life sciences, especially biology. In that context, this essay is concerned with the place of organization within mind and the place of mind within the life sciences, especially biology. There are many possibilities for theories of mind, ranging from noumenal to neural to nihilist (behaviorist), and for most of these, the question of the role for organization therein makes no sense; further, they escape, or are opposed to, any deep tie to biology. Even when some link to biology is acknowledged, as for physicalisms, no inherent notion of organization appears in their development. But this chapter will present a thoroughly organizational conception of mind-as-cognition, anchored in a supportive conception of biology.

5.1 Introduction

The theme of this book is the place of organization in the life sciences, especially biology. In that context, this essay is concerned with the place of organization within mind and the place of mind within the life sciences, especially biology. There are many possibilities for theories of mind, ranging from noumenal to neural to nihilist (behaviorist), and for most of these, the question of the role for organization therein makes no sense; further, they escape, or are opposed to, any deep tie to biology. Even when some link to biology is acknowledged, as for physicalisms, no inherent notion of organization appears in their development. But this chapter will present a thoroughly organizational conception of mind-as-cognition, anchored in a supportive conception of biology.

There are three versions of how something – here, cognition – is bio-organizational, each more stringent than its predecessor. (I) Cognition is best understood from within a bio-cognitive organizational framework. (II) There is a key high-level organizational characterization of cognition. (III) At the core of cognitive function is organization. Here explanation is ultimately dynamical explanation, and these three characterizations of cognition are to be considered as three degrees of explanatory centrality for organization, rather than, for example, as three distinct conceptual kinds (see below).

Consider, in illustration, an unheated pot of fluid on a stove, its liquid molecules moving at random. There is neither ordering nor organizing. Then the stove is used to gently heat the bottom liquid layer. The liquid forms ordered horizontal layers, warmest at the bottom, coolest at the top. Molecular motion remains random horizontally, but the vertical symmetry of random motion is broken, replaced by an ordering of layers by temperature (random molecular energy) that conducts the heat slowly upwards. Finally, as heating increases, rolling boil (or Benard) cells form, vertical and horizontal symmetries are broken and random motion is replaced by a pattern of cells within each of which molecules move circularly, conveying hotter liquid up to the fluid surface and cooler liquid back down to be reheated, their intra-cellular circular motions so arranged horizontally that at each pair of adjacent cell surfaces they are moving in the same direction. The whole manifests moderate order (horizontal) and moderate organization (vertical and horizontal). Then, following the three nested requirements for organization as fundamental, we have the following: (I Benard) The phenomena are indeed best understood from within the molecular organizational framework given above. (II Benard) There is a key high-level organizational characterization of the phenomena as representing a succession of molecular arrangements providing increasing heat transfer capacities. (III Benard) At the core of this succession lies breaking symmetries, partly through ordering (vertical stratification), but with the largest capacity shift achieved through coordinated horizontal and vertical reorganization. Nor is more needed for core understanding; (within limits) it does not matter what the fluid is, nor what the heat source is, or what the pot is made of, the sequence of pot states will recur.

5.2 Characterizing Organization

The intracellular Krebs cycle is a useful model of organization. Its function is to transport energy into the cell and eject waste. It is made up of several molecular steps and produces several products, each combining a specific external input with the current internal chemical to dynamically lead to (produce) an output internal chemical for the next step. It is typically diagrammed as a large cycle with several nested cycles driving off it and ordered around it.¹ Organisms are congeries of such kinds of processes, nested from the subcellular (Krebs cycle) to whole organism (e.g., respiration), all component processes appropriately space-time interrelated.

This is not so different from a motor vehicle engine where all of the many kinds of components are very different from one another (cf. fuel injectors, spark plugs, cam shafts) yet are interrelated in many distinctive ways so that together they perform the transformation of fuel into linear motion. We can think of this as an interrelated structure of sub-functions – fuel injections into cylinders, cylinder heads rising up, sparking the injected fuel, etc. – that together bring about the overall global function. In a clarified ontology, each sub-function is realized as a causal process (one driven by an energy gradient) that takes its start as the function initial condition and moves dynamically to generate the function end condition. In many situations, sub-functions and their realizing dynamical processes may come and go as part of the overall function/process. Immediately after sparking a cylinder head, a large energy gradient forms in it which forces it back along its cylinder shaft. But this gradient only lasts until the fuel is “burnt.” Then another cylinder takes its turn. Similarly, there are many biochemical interactions in molecular biology that only briefly exist while some momentary, but precisely located, function is realized. ln the case of the engine, the constraints that structure and stabilize these processes include the entire engine frame and are much longer lasting than individual cylinder processes. And this is common for current human-made machines. But in molecular biology, it frequently occurs that the whole realizing processes, energy gradients and constraints are ephemeral, changed by equally ephemeral processes of which they are temporarily a part (cf. a seasonally eroding river bank and its flow). This should be understood as normal. The core process organization that grounds cognition relies on just such a structure (see III below).

“Organization” has a narrow and a wide usage. In its narrow usage – n-organization – it means possessing internal, nested correlations of the general sort illustrated above in the Krebs cycle and car engine. In its wide usage, ‘‘organization’’ means no more than ‘‘is in some respect, to some degree, systematic,’’ as in having a well-organized work desk. In this wide sense, one may speak of hierarchical organization though only an ordering by parts and composition is intended, and whether or not internal nested correlations are involved. ‘‘Self-organization’’ as commonly used includes molten iron cooling down to a solid bar (the ion lattice is ‘‘well organized’’), and ordering coins by size through random vibration against varying mesh-widths. In neither case is n-organization part of the output. And in neither case is there any more than the faintest suggestion of a ‘‘self-active’’ process. But both of these examples have new constraints as output. This feature generalizes: self-organization is best conceived as a process leading to the emergence of new constraints, whether or not they produce n-organization and whether or not there is an active self involved (see Hooker, 2011c). Here we are concerned only with n-organization where, as we shall see, it forms a distinctive class of biological conditions.

Ultimately, all n-organization is grounded in dynamical processes, as are all non-organized (a-organizational) states and behavior. N-organized dynamics grades into a-organized dynamics (i.e., plain old dynamics) as the internal processes show decreasing variety, decreasing uniqueness and complexity of collective functions realized and increasing dependence on specific dynamical conditions. But the universality of dynamics is the same in all cases. Two billiard balls colliding show no n-organization but are fully dynamical; the Krebs cycle is strongly n-organizational but each transformation is fully chemo-dynamical. N-organization carries only relationship or form, not quality; quality is carried by dynamics, including the dynamics that grounds relationships or form. This applies to cognitive accounts (e.g., Russell’s electrical charge, Penrose’s intracellular coherent quantum states – Russell, 1927; Penrose, 1989). Here only n-organizational character will be considered.²

N-organization and Order

In terms of interrelations between components, n-organization lies between complete disorder and complete order. Complete disorder is where all component interrelations are random, so that there is no simplifying multi-component pattern which constrains their interrelated behaviors. With completely ordered components, there is a governing pattern, illustrated in soldiers marching in tight formation, or crystals in a uniform lattice, and also distinct from the random collection of its components. With n-organized components, there are also governing patterns, but these can be much more complicated and subtle than the simplicities of complete randomness or complete orderedness (cf. Krebs, engine).³ Wholly random and wholly ordered are poles of zero internal n-organization, all n-organized systems falling somewhere between them. Bennett proposed the notion of logical depth to capture a formal notion of n-organization located along this continuum (Bennett, 1985). Roughly, logical depth is the number of nested correlations within correlations in an entity. This is certainly an important step in the right direction because it places distinctive correlations at the heart of n-organization. But obtaining a satisfactory measure for degree of n-organization is not easily done.⁴ Further exploration lies beyond the scope of this paper.⁵

5.3 N-organization and Bio-cognition

Briefly, Looking Ahead

First, it is argued that a specific kind of n-organization, called autonomy, characterizes all and only living organisms. Autonomy is shown to ground all the major n-organizational aspects of agency. Second, cognitive agency, the main objective here, is in turn shown to be a sub-class of autonomous agents and so ultimately a specific class of n-organized systems. Third, cognitive agency spans a range from elementary to deep problem-solving powers, a range that can be characterized n-organizationally. In sum, autonomy > agency > cognition > deep cognition, with each step along the way, distinctively and strongly characterized n-organizationally. With this framework in mind, let us proceed.

5.3.1 Autonomy, Agency (and Robotics), Auto(self)-directedness, and Anticipation

Autonomy

Our concern in this paper is with the place of n-organization in a biologically centered account of mind. Even so, it is essential to begin with at least one aspect of the wider issue of the place of n-organization within biology generally. N-organization lies at the heart of what an organism is and when we properly understand how that is, we shall have constructed the basis for an n-organizational account of organism minds.

At their most basic, all living things are thermodynamic engines, existing in a far-from-equilibrium condition only maintained by conversion of an input flow of negative entropy (food) to do work and by the export of unutilized material to the environment (wastes). Essential work is of three kinds: (i) the repair or replacement of internal infrastructure, including of any enclosing membrane, and of the capacity for suitable work, (ii) the support of action in the environment, and (iii) the export (elimination) of wastes. This is already an n-organizational arrangement, focused around two cycles, an external interaction cycle with the environment comprising resource extraction and waste elimination and an internal action cycle comprising repair and replacement.

There are various obvious constraints on successful versions of this n-organizational arrangement: (C1) the negative entropy input flows have to arrive in a timely manner, at appropriate places and in appropriate quantities, to sustain all the organism’s processes; (C2) the internal work doable on these flows by the organism must produce sufficient components to fully support the internal repair work, including reproduction of the repair capacities; and (C3) at the same time, their consequent resource exploitation and waste accumulation must be extractable and exportable by the organism at sufficient rates and volumes as to avoid both direct damage to the organism internally and indirect damage via environmental damage. Despite their apparent particularity, these constraints are in fact permissive in form. For instance, it does not matter whether the food-gathering action is largely passive (e.g., a pitcher plant trapping insects) or active (e.g., a dragon fly hunting insects), discriminating (e.g., a koala’s taste for eucalypt leaves) or indiscriminate (e.g., the pitcher plant); it matters only that it satisfies at least the constraints C1–3.⁶

The condition for organism viability is that each cycle is supported and the two cycles so interact as to meet constraints C1–3 above. This n-organized dynamical viability condition is called autonomy.⁷ It picks out all, and plausibly only, living individuals – from cells to multicellular organisms to various multi-organism communities, including many (but by no means all) business firms, cities and nations. There is an issue of how sub-function processes might exactly fit together, each helping to canalize others (e.g., Kauffman’s work-constraint cycles, Kauffman, 2000), to achieve self-reproduction on a sufficiently small scale (contrast engine repair and the whole economy), but in principle some combination of longer-lasting and ephemeral process constraint formation should do it.⁸

The name is appropriate: in autonomous systems, the locus of living process regulation lies more wholly within them than in their environment. Birds use twigs to make nests, but twigs themselves have no tendency to use nests or birds to any purpose. Hence the root sense of autonomy in the traditional social sense. Moreover, there is a richness to the notion that escapes the bare appearance of inter-locked cycles. Autonomous entities have a distinctive wholeness, individuality and perspective in the world derived from the global, interconnected nature of their cycles and the regenerative condition they sustain. This gives rise to achieving (or not) an integrated condition of satisfying autonomy (or not).⁹ Further, when this satisfaction condition is available to the organism itself as some kind of associated signal (e.g., absence of enclosing membrane stress), then the autonomous system has a basic sense of normative requirement. A situation will come to be identified as injurious (reduced integrity), healthy (increased integrity), or neutral, an evaluation that amounts to a distinctive normative perspective. In this manner, autonomous system activities are also willful, anticipative, deliberate, normatively self-evaluated, and adaptive. Such entities are properly treated as genuine agents. Autonomous systems are inherently all of those things.

Agency and Robotics

Meanwhile, let us pause to briefly consider technologies in relation to autonomy and the possibilities of autonomy-based robotics. The dominant difference between biology and technology is, as the petrol vehicle illustrates, that organisms are much more active, responsive and integrated entities than technological systems are, or are often capable of being. A primary difference lies in the inner loop. Vehicles are not self-repairing. Their metabolism is outsourced to repair specialists (mechanics), and from there – via manufacture of spare parts and tools for repairing, the process strengthened by human n-organizational technologies such as pacemakers – to the rest of the economy. Plants do reproduce branches and roots, and both they and animals adaptively alter their bodies in response to environmental interactions, but in animals these alterations are mostly confined to nervous systems, while plants self-maintainingly adapt their bodily forms to support photosynthesis; and both, like vehicles, rely on a larger ecology for the resources to do so, and so on. Pursuing these analogies raises issues concerning how widely distributed, ‘‘socially’’ interlocking and interactively open an agent’s body may be, and conversely how deeply capacity-modifying protheses may be integrated with ‘‘natural’’ agents, and how much does reliance on surrounding ecology for repair differ n-organizationally from reliance on societal economies (cf. notes 9, 10). A second primary difference lies in the external loop: organisms are much more active in responsively regulating their interactions with their environment, and within themselves. This difference is deeply rooted in organism autonomy which provides them a self-orientation to the world that works on integrating many streams of information (perceptual, proprioceptive, affective, etc.), using them to enrich and modify their anticipative interaction models and the directed responses to which it gives rise. Finally, organisms show a wide variety of boundary forms and defenses, from an identifiable exclusionary membrane offering regulated intake of specific nutrients (e.g., gastrointestinal membrane) to socially constructed maintenance of internal community regulation (e.g., through mating roles) and to a highly inter-penetrating film through which DNA may be directly interchanged. These differences are rooted in differences of n-organization.

This autonomy-based characterization of agency meets all the criteria for being deeply n-organizational: (I) Agency is best understood from within a bio-cognitive n-organizational framework, namely, the inter-locking cycles underlying agent autonomy. (II) There is a key high-level n-organizational characterization of agency, namely, as expressing autonomy. (III) And since every capacity of agency is based on autonomy, whose core is n-organizational, the core of agency is n-organizational. The distinctive n-organizational character of life penetrates deeply into its nature, into universal roots constituting agency. And it will be on this basis that any account of mind as n-organizational will be built and find its place in biology.

Meanwhile, there remain 3 + 2 robotics issues. (A) How might the constitution of an integrated internal perspective be achieved, if at all? What role has autonomous n-organization in the construction of robotic focused and responsive bodies? What might their perspective on the world be? (B) How does the manufacture of tools by tools and commodities by commodities proceed and how, if at all, does it include all elements (And how is it shown that the manifest tool improvement that does in historical fact occur within it, can actually occur within it?) (C) What is the biology and sociology of boundaries, how can these be constructed robotically and with what consequences for internal n-organization? Each of these presents a deep and subtle problem. They are left for the reader to consider, as is their impact on the n-organizational character of these aspects of natural and artificial existence. Only after these issues have been addressed will there be a proper platform for addressing the issue. (D) What are the limits to autonomy? And then (E) can there be a truly autonomous robotics?¹⁰

Returning to the main argument, its overall structure is as follows. Self-directedness and anticipativeness are two fundamental cognitive capacities harbored by autonomy. Mutually supporting one another, these capacities form the central cognitive process of self-directed anticipative learning (SDAL). SDAL in turn provides the foundation of the deepest, most powerful forms of problem-solving, that is, of cognition, and of tracking, that is, of intentionality. Thus, intention and cognition are provided their common n-organizational root.

Auto(self)-Directedness

Auto- or self-directedness [the latter, more common, version will be used] is the capacity to self-modify interaction in the light of its evaluation by the directing organism. Changing behavior to acquire newly available food (e.g., spring flower nectar) is one example, and changing behavior to manage pain is another. Such sensitive, conditionalized attention forms the intertwined root capacities of intention and cognition (cf. Christensen & Hooker, 2004). A mosquito has one known such process (whether or not to initiate search for a blood host – Klowden, 1995), and a mammal has a vast number of such conditionalizing processes, especially within its motor regulatory system. Cycles (the n-organized aspect) of signaling and initiating specific actions within the external interaction cycle, and evaluating their outcomes against autonomy support through the internal interaction cycle, provide the basic autonomy n-organization with strong outcome-led self-directedness. In appropriate context, something about the direction of value increase (i.e., autonomy support) is also provided (e.g., by testing small departures in various directions from the present setting to see which is more rewarding). In more sophisticated form, self-direction allows the interplay of multiple evaluative signals, combining, compromising and conditionalizing them when arriving at which values are appropriate for guiding action in the context, recognizing corresponding multiple streams of information as relevant to those decisions, and in that light follows their integration for regulation of decision-making. The more mutually convergent guiding values and streams of information the learner has about performance, the more effective its actions can be. Initially, guidance will be limited because of learner ignorance, while at the concluding stage, information will have been sufficiently enriched, focused and integrated into the interaction cycles as to allow the learner to converge on a solution.

Anticipation

There is another, closely related, feature that the mosquito and the mammal share (very unequally): anticipation. Reflex and random actions aside, every action anticipates its outcome. At its most primitive, anticipation is the forming (learning) of a simple association between current features and an outcome of an action. The bee dance anticipates re-locating ephemeral nectar supplies as outcome; it would not be attended to unless that outcome and its attendant resource availability were frequently enough the consequence of the dance.¹¹

Anticipative learning is where the organism learns to anticipate a goal achievement by employing an action sequence, thus associating receiving goal satisfaction with doing an action sequence. Elementary associative learning such as neural conditioning provides the simplest anticipative associations. More sophisticated versions of this process can be elaborated as learning capacities widen. For instance, though much more sophisticated than the mosquito, the cheetah swinging rightwards chasing a dodging gazelle with a right-swing bias is still doing so anticipating a desirable outcome (a kill). But with the cheetah, all the mammalian apparatus of planning ahead, tracking trajectories for oneself and others and so on is put to use managing these interactions fluidly and at high speed. The cheetah’s many associations – approach downwind, remain camouflaged where possible, maintain prey separation from herd and so on – have come to be integrated in richly associated models (here of the hunt). Bringing all these capacities together, in n-organizationally mutually supportive ways, provides the close attentiveness to problem-solving that is the core of intentional cognition.

Self-directed Anticipative Learning [SDAL]

The combination of self-directedness and anticipative action provides the basis of fluid self-steering. An action is undertaken in anticipation of achieving a goal; if it does so, the anticipation is entrenched, and if it does not, the action may be repeated, extended or modified, at the actor’s self-direction. In this way, the actor steers itself through a process of learning its environment. The capacity this invests in its agents is adaptability. The ultimate goal of external adaptation is internal regulation, i.e., to be able regulate the operation of the twin autonomy cycles so as to continue to satisfy autonomy, in the environmental circumstances obtaining. However, in a dynamic environment, where creatures are constantly changing (e.g., their current location and posture), often across many fronts and on many time scales, detailed adaptation is momentary and only approximated. Instead, it is necessary to be adaptable, able to adapt once useful adaptations as new conditions emerge (run from a predator, switch diet, migrate, etc.) There are limited physiological adaptabilities, most subconscious and of fixed operation (e.g., callous formation, switching to burning visceral fat to extend flight in extremity). But the largest, most variable and most rapidly adaptable are the behavioral adaptabilities, from singing to flying to technology construction, regulated by the central nervous system and largely expressed through the motor system.

These latter features (largest range, most variable, most rapidly adaptable) do not in themselves constitute more than small augmentations of cognitive power. Fluid adaptation ranges from the superficial to the deep, and these add finesse to the superficial capacity. That the flatworm withdraws into the shadows in a larger range of ways and circumstances, more variably, and faster, when a light is shone on it, does not modify its simplicity, or its fixity, of response. Superficial adaptation offers fixed information channels and evaluation routines that provide only first-order fixed responses to changing situations, the whole working off an n-organized algorithm without the need of higher-order regulation, something that fairly cheap route planners and guided missiles, along with rafts of insects, worms and others, can provide. Moving toward greater capacity involves increasing numbers of conditionalizations, supporting increasing spread and discrimination of judgment. Though always useful, this level of fluid but fixed regulation cannot surmount significant shocks such as failure to recognize anticipated response sequences, or interaction dynamics altering mid-action.

Beyond this impasse lies the introduction of increased layers of higher-order conditionalizations, offering increasing orders of responsiveness and increasing integration of responses. Sufficiently developed, such higher-order, integrated judgment formation underlies powerful new dimensions to problem-solving, for instance, the capacity to respond to a ‘‘mis-match’’ signal as indicating, not merely a new trial in response, but a change in investigative methods used. Consider discovering through a mis-match signal (e.g., unexpected viral outbreaks) that the present testing method has an unexpected high false-negative rate (say in pharyngeal swab testing for a viral infection), requiring a change in testing method to achieve greater reliability in estimates of infection rates and hence in demand for healthcare resources, and so on. As well as method change, consider also bringing about reformulations of the problem to hand (‘‘It’s not the measuring process, but it’s the modeling of sub-population interactions we are using’’), changed criteria for successful outcomes (“predictions of infection breakout locations and frequencies accurate to within 10%”), changed external constraints framing the problem (‘‘rural sub-populations are much more constrained by travel times’’) and changed criteria for “cleaned” data supporting these judgments (‘‘estimates of false positives as well as of false negatives are required’’). As will appear, supporting the integration of these features will provide deep fluid adaptation, or deep adaptability, the mark of truly human intelligence. (In this respect, we are a long way yet from deeply intelligent robotics.)

To see how these features work together, consider a detective conducting a murder investigation. She uses clues from the murder scene to build an initial proposed profile of the suspect and then uses this profile to focus the direction and methods of the investigation. Lipstick on a glass suggests a crime of passion, with the suspect female, in a personal or sex worker relationship to the victim. The profile tells the detective what the murderer may be like and what characteristic types of clues to pursue. For a crime of passion, look for further personal effects – special clothes, whips or other ‘‘technologies’’ in producing sexual effects, etc. Look too for places nearby, possibly frequented for romantic assignations, a romantic bar, a brothel, etc. The chosen profile in turn sets new intermediate goals, for example, narrow down the nearby places frequented, eliminate or reduce the likelihood of the suspect being a male cross-dresser, but conversely try to obtain an estimate of how many women might be involved. If the chosen profile is at least partially accurate, and with a little luck, the modified investigation will uncover further evidence that in turn further refines the search process, ultimately culminating in the capture of the murderer, and resolving the nature of the investigation.

But such searches are not fixed; a good detective will have in mind other possible profiles awaiting supporting evidence. Further search of the murder site, for example, may uncover a gambling note for a substantial sum. This turns attention to enforcing debt default as the kind of crime involved. The lipstick cue does not fit comfortably into this version; the culprit is more likely a male, with a history of criminal activity and likely enforcer violence. This profile redirects the search from sexual partners to gambling associates and perhaps money laundering and the like. From this point of view, the lipstick is mis-directing; perhaps it belonged to an attempt to persuade the victim to settle his debts, or was indeed worn by a cross-dresser, but just as a personal quirk, irrelevant to the financial issues at stake, or planted to ‘‘throw the investigation off the scent.’’

It is the interplay between the discovery of clues, the construction of a suspect profile and subsequent modification of the investigation that makes the process self-directing. It is powerful self-direction because it encompasses re-thinking the nature of the investigation (here from sex to gambling), contextual assumptions (here from assignations to debt collections), data (lipstick from evidence of lover’s presence to irrelevance), and solution types (from identification of passionate conflict, murder process and culprit motive to identification of debt association and assassin presence and actions). As an organism interacts in an SDAL process, its improving anticipative models and model-based interaction processes allow it to (a) improve its recognition of relevant information, (b) perform more focused activity, (c) evaluate its performance more relevantly and precisely, and (d) learn about its problem domain more effectively. Indeed, in this setting, error itself can be a rich source of context-sensitive information that can be used to further refine these four features.¹² The richer the system’s anticipative and norm structures are, the more directed its learning can be, and the more potential there is that learning will improve the system’s capacity to form successful anticipative models of interaction. To this the detective adds an additional kind of learning, higher-order learning about the entire domain of murders. It is the capacity to learn across many such investigations what sorts of profiles there are; what are their general features and rare exceptions; what their associated kinds of investigatory methods, timetables and costs; and so on and to recognize when there is more to learn and how to be alert to doing it that makes the process such a powerful problem-solving tool.

When that kind of ‘‘double-loop’’ learning occurs, the detective is both learning what works, or not, in the immediate investigation to hand and at the same time using that experience to improve general knowledge of detecting murders, and crimes more generally – knowledge that will in turn be used to improve the next investigation. In short, by learning a higher-order characterization of the problem class (murder investigations), she will have been learning how to learn about investigations in that domain while learning how to solve specific problems to hand. Just this is the fundamental bootstrap required for all learning to be improvable. It forms the key to understanding the n-organization, and thence the general power, of the learning process. Indeed, this process allows the rational resolution of initially ill-defined problems, problems whose formulation and structure are vague, gappy or ambiguous, or tacitly internally conflicted, or whose valid methods are unsettled, like how to detect or marry well, or validly test a scientific theory in a new domain. Such problems of necessity lie at the root of every new problem-solving domain.

It is possible to synthesize a model process for such learning-how-to-learn-while-learning processes. Each of the five foci or nodes of learning identified above (method, problem formulation, solution formulation, constraints, data) are represented. As each specific learning process is under gone, attention shifts from one node to another, or to several nodes in parallel, as the potential consequences of experimenting with alternatives are explored. (Cf. the different investigations formed by the detective’s various crime profiles.) Eventually (and with some luck), the investigations are reduced to one, the one that resolves the core detecting problem. Although all investigations share the same n-organizational form, the non-organizational features play their decision structuring and making roles alongside them and varying from incidental to central. A measure of their importance is the degree to which they must be appealed to at each choice point. For this reason, there is no specifiable model, let alone algorithm, for the order in which nodes are visited, nor for what changes are consequently made, nor for how these changes in turn spread across the nodes, nor for what kinds of compromises are made in reaching for an enriched solution, and so on.¹³

Yet the model does capture the fundamental n-organization of kinds of actions that deliberative problem-solving centrally involves. In its lowest form, this n-organization is expressed in the cyclic processes of specific trial-and-error interrogations. Moving to higher-order organization, these n-organized trial-and-error processes are nested within sharings of information about how to coordinate the findings from several such trials covering these kinds of crimes. Every cheetah hunt, and every detecting, is unique in its qualitative details but they are also all the same as n-organized hunting processes. In particular, they all share the higher-order prospects of reformulating the problem and/or the solution, changing and nested again inside more general formulations of investigating crimes of these general kinds. This tri-layer of nested cyclings is of the same general form as the Krebs cycle (above), but here its n-organizational depth is much greater because, for example, at each of its nodes, it stores structured content about the domain related to that node, and stores cross-nodal interrelations pertinent to the domain involved, both contents increasing their richness as problem-solving multiply, none of which the Krebs cycle has available. In this enriched form, the problem-solving model has deeply illuminated the 30 years of research into the linguistic capacities of apes, even how (pace Kuhn) rational deliberative problem-solving can proceed through scientific revolutions.¹⁴

5.4 In Conclusion

It remains to reiterate that this model of problem-solving is primarily n-organizational. No matter the domain concerned, this moderately n-organized, moderately ordered process successfully models the general problem-solving process. The underlying sense of agency on which the problem-solving SDAL process is built is fundamentally n-organizational, satisfying the three criteria for being essentially n-organizational: (I) best understood from within a bio-cognitive n-organizational framework, (II) has a key high-level n-organizational characterization, and (III) its core is n-organizational. The distinctive n-organizational character of life penetrates deeply into its nature, into its universal roots constituting agency. And now it is on this basis that the roots of cognition have also been revealed to be in essence n-organizational. (I) Problem-solving is best understood from within a bio-cognitive n-organizational framework, here that of autonomy-based bio-agency, with its distinctive accounts of identity and normativity. (II) There is a key high-level n-organizational characterization of problem-solving, namely, that of the general SDAL problem-solving process model. (III) The core of problem-solving is n-organizational because it lies within the improvable, enrichable tri-nested cyclicities of the general problem-solving process model. Goal-pursuit is an inherent, if moderate, n-organized process; it marshals the steering sub-processes – anticipation and self-directedness – to orient to the goal and to explore self-improving ways to move toward it. SDAL, equipped with higher-order regulation, is inherently this n-organization. Global-level n-organization is emphasized by the steering processes in SDAL, which are typically higher order. This completes the n-organizational ‘‘golden thread’’ running throughout biology, ultimately integrating mind into living being.

Such n-organizational principles evidently have but small extension beyond life to the cosmos at large. While the inanimate world has n-organization – wherever ‘‘mechanical’’ cyclicities operate, for instance in rolling boil formation (Introduction: Benard cell) – n-organization evidently does not lie deep throughout the cosmos as it does throughout biology. The inanimate makes more use of order than n-organization. No doubt this reflects the simplicity of orderedness and the priority in time of the inanimate world with the emergence of life within it. This makes living n-organization, autonomy, the more remarkable.¹⁵

The details of the general problem-solving process will vary across subject matters. The golden thread of n-organization abstracts from these differences to locate a fundamental n-organizational category: life. It is thinkable that it might not have been so. Understanding how that n-organizational category is possible will involve tracing it back to the fundamental qualities as we know them, the quantum and relativistic qualities: mass, spin, charge, and so on, along with those that structure irreversible thermodynamics. There is at present no neat accepted story here, and the problems are so deep and unresolved as to make it thinkable that there is none for finite mortals to have. The complications rise further if those qualities associated with mind, the perceptual and emotional qualities, are included. It is always possible to try for a purely process account of these, or for a more n-organizational one, though how complete they can be also currently remains open.¹⁶