Tag: Francisco Varela

Varela’s modes of explanation and the teleonomic

I’m now diving deep into Francisco Varela’s Principles of Biological Autonomy (1979). Chapter 8 draws on his paper with Maturana, “Mechanism and biological explanation” (1972) (html). Chapter 9 draws heavily from his paper, “Describing the Logic of the Living: adequacies and limitations of the idea of autopoiesis” (1978) (html).

I am finding this work very enlightening. Somehow it bridges between my interests in philosophy of science right into my current work on privacy by design. I think I will find a way to work this into my dissertation after all.

Varela has a theory of different modes of explanation of phenomena.

One form of explanation is operational explanation. The categories used in these explanations are assumed to be components in the system that generated the phenomena. The components are related to each other in a causal and lawful (nomic) way. These explanations are valued by science because they are designed so that observers can best predict and control the phenomena under study. This corresponds roughly to what Habermas identifies as technical knowledge in Knowledge and Human Interests. In an operational explanation, the ideas of purpose or function have no explanatory value; rather the observer is free to employ the system for whatever purpose he or she wishes.

Another form of explanation is symbolic explanation, which is a more subtle and difficulty idea. It is perhaps better associated with phenomenology and social scientific methods that build on it, such as ethnomethodology. Symbolic explanations, Varela argues, are complementary to operational explanations and are necessary for a complete description of “living phenomenology”, which I believe Varela imagines as a kind of observer-inclusive science of biology.

To build up to his idea of the symbolic explanation, Varela first discusses an earlier form of explanation, now out of fashion: teleological explanation. Teleological explanations do not support manipulation, but rather “understanding, communication of intelligible perspective in regard to a phenomenal domain”. Understanding the “what for” of a phenomenon, what its purpose is, does not tell you how to control the phenomenon. While it may help regulate ones expectations, Varela does not see this as its primary purpose. Communicability motivates teleological explanation. This resonates with Habermas’s idea of hermeneutic knowledge, what is accomplished through intersubjective understanding.

Varela does not see these modes of explanation as exclusive. Operational explanations assume that “phenomena occur through a network of nomic (lawlike) relationships that follow one another. In the symbolic, communicative explanation the fundamental assumption is that phenomena occur through a certain order or pattern, but the fundamental focus of attention is on certain moments of such an order, relative to the inquiring community.” But these modes of explanation are fundamentally compatible.

“If we can provide a nomic basis to a phenomenon, an operational description, then a teleological explanation only consists of putting in parenthesis or conceptually abbreviating the intermediate steps of a chain of causal events, and concentrating on those patterns that are particularly interesting to the inquiring community. Accordingly, Pittendrich introduced the term teleonomic to designate those teleological explanations that assume a nomic structure in the phenomena, but choose to ignore intermediate steps in order to concentrate on certain events (Ayala, 1970). Such teleologic explanations introduce finalistic terms in an explanation while assuming their dependence in some nomic network, hence the name teleo-nomic.”

A symbolic explanation that is consistent with operational theory, therefore, is a teleonomic explanation: it chooses to ignore some of the operations in order to focus on relationships that are important to the observer. There are coherent patterns of behavior which the observer chooses to pay attention to. Varela does not use the word ‘abstraction’, as a computer scientist I am tempted to. But Varela’s domains of interest, however, are complex physical systems often represented as dynamic systems, not the kind of well-defined chains of logical operations familiar from computer programming. In fact, one of the upshots of Varela’s theory of the symbolic explanation is a criticism of naive uses of “information” in causal explanations that are typical of computer scientists.

“This is typical in computer science and systems engineering, where information and information processing are in the same category as matter and energy. This attitude has its roots in the fact that systems ideas and cybernetics grew in a technological atmosphere that acknowledged the insufficiency of the purely causalistic paradigm (who would think of handling a computer through the field equations of thousands of integrated circuits?), but had no awareness of the need to make explicit the change in perspective taken by the inquiring community. To the extent that the engineering field is prescriptive (by design), this kind of epistemological blunder is still workable. However, it becomes unbearable and useless when exported from the domain of prescription to that of description of natural systems, in living systems and human affairs.”

This form of critique makes its way into a criticism of artificial intelligence by Winograd and Flores, presumabley through the Chilean connection.

artificial life, artificial intelligence, artificial society, artificial morality

“Everyone” “knows” what artificial intelligence is and isn’t and why it is and isn’t a transformative thing happening in society and technology and industry right now.

But the fact is that most of what “we” “call” artificial intelligence is really just increasingly sophisticated ways of solving a single class of problems: optimization.

Essentially what’s happened in AI is that all empirical inference problems can be modeled as Bayesian problems, which are then solved using variational inference methods, which are essentially just turning the Bayesian statistic problem into a solvable form of an optimization problem, and solving it.

Advances in optimization have greatly expanded the number of things computers can accomplish as part of a weak AI research agenda.

Frequently these remarkable successes in Weak AI are confused with an impending revolution in what used to be called Strong AI but which now is more frequently called Artificial General Intelligence, or AGI.

Recent interest in AGI has spurred a lot of interesting research. How could it not be interesting? It is also, for me, extraordinarily frustrating research because I find the philosophical precommitments of most AGI researchers baffling.

One insight that I wish made its way more frequently into discussions of AGI is an insight made by the late Francisco Varela, who argued that you can’t really solve the problem of artificial intelligence until you have solved the problem of artificial life. This is for the simple reason that only living things are really intelligent in anything but the weak sense of being capable of optimization.

Once being alive is taken as a precondition for being intelligent, the problem of understanding AGI implicates a profound and fascinating problem of understanding the mathematical foundations of life. This is a really amazing research problem that for some reason is never ever discussed by anybody.

Let’s assume it’s possible to solve this problem in a satisfactory way. That’s a big If!

Then a theory of artificial general intelligence should be able to show how some artificial living organisms are and others are not intelligent. I suppose what’s most significant here is the shift in thinking of AI in terms of “agents”, a term so generic as to be perhaps at the end of the day meaningless, to thinking of AI in terms of “organisms”, which suggests a much richer set of preconditions.

I have similar grief over contemporary discussion of machine ethics. This is a field with fascinating, profound potential. But much of what machine ethics boils down to today are trolley problems, which are as insipid as they are troublingly intractable. There’s other, better machine ethics research out there, but I’ve yet to see something that really speaks to properly defining the problem, let alone solving it.

This is perhaps because for a machine to truly be ethical, as opposed to just being designed and deployed ethically, it must have moral agency. I don’t mean this in some bogus early Latourian sense of “wouldn’t it be fun if we pretended seatbelts were little gnomes clinging to our seats” but in an actual sense of participating in moral life. There’s a good case to be made that the latter is not something easily reducible to decontextualized action or function, but rather has to do with how own participates more broadly in social life.

I suppose this is a rather substantive metaethical claim to be making. It may be one that’s at odds with common ideological trainings in Anglophone countries where it’s relatively popular to discuss AGI as a research problem. It has more in common, intellectually and philosophically, with continental philosophy than analytic philosophy, whereas “artificial intelligence” research is in many ways a product of the latter. This perhaps explains why these two fields are today rather disjoint.

Nevertheless, I’d happily make the case that the continental tradition has developed a richer and more interesting ethical tradition than what analytic philosophy has given us. Among other reasons this is because of how it is able to situated ethics as a function of a more broadly understood social and political life.

I postulate that what is characteristic of social and political life is that it involves the interaction of many intelligent organisms. Which of course means that to truly understand this form of life and how one might recreate it artificially, one must understand artificial intelligence and, transitively, artificial life.

Only one artificial society is sufficiently well-understood could we then approach the problem of artificial morality, or how to create machines that truly act according to moral or ethical ideals.

second-order cybernetics

The mathematical foundations of modern information technology are:

  • The logic of computation and complexity, developed by Turing, Church, and others. These mathematics specify the nature and limits of the algorithm.
  • The mathematics of probability and, by extension, information theory. These specify the conditions and limitations of inference from evidence, and the conditions and limits of communication.

Since the discovery of these mathematical truths and their myriad application, there have been those that have recognized that these truths apply both to physical objects, such as natural life and artificial technology, and also to lived experience, mental concepts, and social life. Humanity and nature obey the same discoverable, mathematical logic. This allowed for a vision of a unified science of communication and control: cybernetics.

There have been many intellectual resistance to these facts. One of the most cogent is Understanding Computers and Cognition, by Terry Winograd and Fernando Flores. Terry Winograd is the AI professor who advised the founders of Google. His credentials are beyond question. And so the fact that he coauthored a critique of “rationalist” artificial intelligence with Fernando Flores, Chilean entrepreneur, politician, and philosophy PhD , is significant. In this book, the two authors base their critique of AI on the work of Humberto Maturana, a second-order cyberneticist who believed that life’s organization and phenomenology could be explained by a resonance between organism and environment, structural coupling. Theories of artificial intelligence are incomplete when not embedded in a more comprehensive theory of the logic of life.

I’ve begun studying this logic, which was laid out by Francisco Varela in 1979. Notably, like the other cybernetic logics, it is an account of both physical and phenomenological aspects of life. Significantly Varela claims that his work is a foundation for an observer-inclusive science, which addresses some of the paradoxes of the physicist’s conception of the universe and humanity’s place in it.

My hunch is that these principles can be applied to social scientific phenomena as well, as organizations are just organisms bigger than us. This is a rather strong claim and difficult to test. However, it seems to me after years of study the necessary conclusion of available theory. It also seems consistent with recent trends in economics towards complexity and institutional economics, and the intuition that’s now rather widespread that the economy functions as a complex ecosystem.

This would be a victory for science if we could only formalize these intuitions well enough to either make these theories testable, or to be so communicable as to be recognized as ‘proved’ by any with the wherewithal to study it.

autonomy and immune systems

Somewhat disillusioned lately with the inflated discourse on “Artificial Intelligence” and trying to get a grip on the problem of “collective intelligence” with others in the Superintelligence and the Social Sciences seminar this semester, I’ve been following a lead (proposed by Julian Jonker) that perhaps the key idea at stake is not intelligence, but autonomy.

I was delighted when searching around for material on this to discover Bourgine and Varela’s “Towards a Practice of Autonomous Systems” (pdf link) (1992). Francisco Varela is one of my favorite thinkers, though he is a bit fringe on account of being both Chilean and unafraid of integrating Buddhism into his scholarly work.

The key point of the linked paper is that for a system (such as a living organism, but we might extend the idea to a sociotechnical system like an institution or any other “agent” like an AI) to be autonomous, it has to have a kind of operational closure over time–meaning not that it is closed to interaction, but that its internal states progress through some logical space–and that it must maintain its state within a domain of “viability”.

Though essentially a truism, I find it a simple way of thinking about what it means for a system to preserve itself over time. What we gain from this organic view of autonomy (Varela was a biologist) is an appreciation of the fact that an agent needs to adapt simply in order to survive, let alone to act strategically or reproduce itself.

Bourgine and Varela point out three separate adaptive systems to most living organisms:

  • Cognition. Information processing that determines the behavior of the system relative to its environment. It adapts to new stimuli and environmental conditions.
  • Genetics. Information processing that determines the overall structure of the agent. It adapts through reproduction and natural selection.
  • The Immune system. Information processing to identify invasive micro-agents that would threaten the integrity of the overall agent. It creates internal antibodies to shut down internal threats.

Sean O Nuallain has proposed that ones sense of personal self is best thought of as a kind of immune system. We establish a barrier between ourselves and the world in order to maintain a cogent and healthy sense of identity. One could argue that to have an identity at all is to have a system of identifying what is external to it and rejecting it. Compare this with psychological ideas of ego maintenance and Jungian confrontations with “the Shadow”.

At an social organizational level, we can speculate that there is still an immune function at work. Left and right wing ideologies alike have cultural “antibodies” to quickly shut down expressions of ideas that pattern match to what might be an intellectual threat. Academic disciplines have to enforce what can be said within them so that their underlying theoretical assumptions and methodological commitments are not upset. Sociotechnical “cybersecurity” may be thought of as a kind of immune system. And so on.

Perhaps the most valuable use of the “immune system” metaphor is that it identifies a mid-range level of adaptivity that can be truly subconscious, given whatever mode of “consciousness” you are inclined to point to. Social and psychological functions of rejection are in a sense a condition for higher-level cognition. At the same time, this pattern of rejection means that some information cannot be integrated materially; it must be integrated, if at all, through the narrow lens of the senses. At an organizational or societal level, individual action may be rejected because of its disruptive effect on the total system, especially if the system has official organs for accomplishing more or less the same thing.