The Autonomy of Life

A short article by Keith Farnsworth Jan 2019 Still under construction

The defining feature of life might turn out to be its autonomy.

Life differs from everything else we know in the universe in that we cannot (yet) explain it without reference to ‘agency’, or purpose. At the heart of all explanations of life - self-maintaining, self-propagating, adapting and learning - we ascribe a reason and role to every part or system composing life. The most parsimonious understanding of this reason is life itself: we might say that all life exists to maintain and reproduce life (see autopoiesis). But we would not feel the need to say such a thing about any non-living system: no galaxy or rock or river exists to maintain itself, nor for any other purpose. Not even fire, or whirlpools need such a statement, though they do maintain themselves, using resources from around them. But fires just happen: they follow the laws of chemistry with the laws of physics right behind: directing cause and effect as mere “happenings”, blindly, without need and without purpose. We do not need to refer to what fires “do” in order to understand them. It is enough to understand the chemical reaction of exothermic oxidation, which in the right circumstances, just happens. So why, when a bacterium swims towards a sugar concentration, must we refer to what the bacterium is doing and what functions the internal mechanisms of its little body are given to enable this.
The reason is that to properly account for them, we need to recognise living organisms as whole things, with a definite boundary and a causal relationship joining every part together into a meaningful whole, which on the level of the organism, acts with unity to give a definite identity to the system that is that organism. So bacteria swim, birds listen, plants compete for water and nematodes attack their tiny prey, whilst galaxies and rivers just are; and their actions and reactions just happen.  It is the very fact of being an organised whole that makes an organism something which does as well as is done to. What lies behind this is the very curious phenomenon of primary cause apparently arising from within the organism, rather than it being merely a link on the chain of cause and effect that leads ultimately back to the 'big bang'. This curious phenomenon is autonomy (the most advanced form of which is free will).

What really is Autonomy?

Let us start with Sharov's (2010) definition of an active agent: anything that can make decisions and act in the physical world. Strictly speaking, this means without exogenous control. The Chinese robot spacecraft that landed on the dark side of the moon (2019) had to be autonomous to some extent because there can be no communications with it over there. These days it is not hard to imagine an autonomous robot in the sense of making decisions for itself without exogenous  control. It certainly does not have to be following a pre-determined algorithm of decision making in which anticipated situations and decisions have been programmed into an algorithm for it to follow (as human-made cybernetic systems used to be). No, using 'machine learning', it could be quite an independent 'thinker'. Indeed, as demonstrated in Farnsworth (2017) a robot of this kind could even gain a degree of 'free will'. 

In that paper, I provided the following definition.
 An agent has 'free will' if all of the following are jointly true:
•    FW1: there exists a definite entity to which free-will may (or may not) be attributed;
•    FW2: there are viable alternative actions for the entity to select from;
•    FW3: it is not constrained in the exercising of two or more of the alternatives;
•    FW4: its “will” is generated by non-random process internal to it;
•    FW5: in similar circumstances, it may act otherwise according to a different internally generated “will’.

In this definition the term “will” means an intentional plan which is jointly determined by a “goal” and information about the state (including its history) of the entity (internal) and (usually, but not necessarily) its environment. The term “goal” here means a definite objective that is set and maintained internally.

The list of criteria (FW1-FW5) was chosen to address the main features that most philosophers have thought important (though they do not necessarily agree with one another about what is important and some philosophers would leave out some items of the list). FW2 and FW5 are intended to examine the effect of determinism and FW4 represents the “source arguments” for and against free will, whist FW3 ensures freedom in the most obvious (superficial) sense. Only one of the list (FW1) is not usually included in any philosophical discussion of free will, perhaps because it is usually considered to be self evident, but it will play an important role here.

The implication of agency is that a system to which it is attributed acts in a way that is systematic (not random) and such that the state (and maybe history of states) of the system (agent) is one of the determinants of the system’s behaviour. More specifically, the next state of the system is not random, not wholly determined by exogenous control, nor intrinsic to its structure (as in clockwork), but is at least partly determined by its present and (optionally) one or more of its previous states. The proximate cause of an action taken by an agent with agency is identified as its ‘will’. This proximate cause is not merely mechanism [44], it is the result of information with causal power rather than just deterministic effective cause (see discussion of causes).


How can a thing be the ultimate cause of its actions?

Identifying ‘it’ : The Kantian Whole

The idea that living organisms require an explanation of agency because they are recognised as wholes’ (in which component parts logically seem to have a role), was first formalised by Emanuel Kant in his “Critique of Teleological Judgement” (well described in Ginsborg, 2006). As Kant put it: for something to be judged as a natural end "it is required that its parts altogether reciprocally produce one another, as far as both their form and combination is concerned, and thus produce a whole out of their own causality". A system composed of parts which in turn owe their existence to that of the system is accordingly termed a ‘Kantian whole’ by Stuart Kauffman (Kauffman and Clayton 2006) and it seems a pity that his terminology has not caught on yet. The difference between living and non-living things in this context is that among the non-living, our identifying of separate objects is a subjective convenience: in fact they have no separate identity at all, they are just identifiable components of the whole of the universe. When an atom vibrates in one part of the universe, the effects propagate throughout it for all time and nothing happens without a prior cause elsewhere and the cause of everything is everywhere.

In contrast, living things are logically separate precisely because they are necessarily organised wholes, in the sense defined by Kant. The whole is an autopoietic system: it makes and maintains the parts. The parts constitute the whole and by their collective interaction give rise to emergent properties that belong to the whole and not to them. The whole is bounded by a tegument which has selective capabilities, filtering what is necessary from the environment, rejecting the rest and expelling waste. The components collectively form an autocatalytic set in which each part catalyses the production of others and no more. In this the organism is self-referencing, for to make each part, one needs every other part, but each of these needs the part we started with. This state of affairs constitutes task closure (equivalently - operational closure) in which every part is the cause of every other, but only in the context of the whole and this is the cardinal sign of a Kantian whole. The self-referencing among the parts isolates them from their interconnection with the rest of the universe. Parts of the whole are related to one another in cause and effect, rather than the whole universe. This creates the (apparent) autonomy of the whole, for what happens within the organism is determined by the organism, specifically the organised happenings of the interacting parts, without (direct) reference to the rest of the universe. The organised whole acts in accordance with its autopoietic nature - it modifies its immediate environment (especially internally) to maintain its integrity. This is not a decision, of course, it is an inevitable consequence of its organisational structure. A bacterium is just an organised set of chemical reactions, but it is one that makes, maintains and reproduces itself. It does this consistently in a range of environmental circumstances, so it is to this extent independent of the nature of the universe surrounding it. That makes it crucially different from any non-living system, for which any change in circumstances results in a change of internal process (and perhaps composition). For the organism, there is an ‘it’ with a separate identity and some degree of autonomy of action.

That said, the Kantian whole concept is neither precise enough, nor accurate enough for

I am not saying, as Kant and others have said, that the parts exist for the whole and the whole for the parts. I talk of the manner in which the molecular process interconnect with each other so that a living system exists as a totality that appears to an observer as if the parts existed for the whole and the whole for the parts -- which is not the case. The components of any system exist as local entities only in relations of contiguity with other components, and any relation of the parts to the whole established by the observer as a metaphor for his or her understanding has no operational presence - Humberto Maturana Romesin.



Organisational and Causal Closure

This box explains the (at first rather opaque) idea of closure and its different kinds.
This may seem a little arcane, but it turns out to be tremendously important for precisely working with causality and autonomy. A Kantian whole may be more precisely defined as a system with transitive causal closure. To understand this, we will start with the definition of closure in general, this leads directly to operational closure, with which we explain transitive closure and then identify organisational  closure, finishing with causal closure.


Closure: is a mathematical concept applying to sets of relations. In general, a set ‘has closure under an operation’ if performance of that operation on members of the set always produces a member of that same set: this is the general definition of operational closure. If the operation is relational (e.g. A>B), then the system has relational closure if it has operational closure under the relational operator.

It can be applied in many relevant systems, for example if X is a set of chemical species and R a set of reactions, then if for every possible reaction in R among members of X, the products are always also members of X, then the set X is closed under R (this is one of the prerequisites of autocatalytic sets).

This was used in the ‘Closure Thesis’ underpinning Maturana and Varela's theory of autopoiesis: every autonomous system is operationally closed. Quoting Varela [25] “A domain K has closure if all the operations defined in it remain with the same domain. The operation of the system has therefore closure, if the results of its action remain within the system itself".


Transitive closure. In maths, if A->B and B->C then A B C is transitive only if A->C.

Organisational closure. According to Heylighen [73], “In cybernetics, a system is organisationally closed if its internal processes produce its own organisation” (my emphasis). This was used in the ‘Closure Thesis’ underpinning Maturana and Varela's theory of autopoiesis: every autonomous system is operationally closed. Quoting Varela [25] “A domain K has closure if all the operations defined in it remain with the same domain. The operation of the system has therefore closure, if the results of its action remain within the system itself". More loosely, Vernon et al. [54] stated that “the term operational closure is appropriate when one wants to identify any system that is identified by an observer to be self-contained and parametrically coupled with its environment but not controlled by the environment. On the other hand, organizational closure characterizes an operationally-closed system that exhibits some form of self-production or self-construction” [111].

Causal closure.
Rosen 1991 used the Aristotelean language to describe the special case of causal closure, calling it “closure to efficient cause”. This is an operational closure in which the operation is causal. This idea was formalised by a mereological argument (mereology deals with parts and wholes with formal logic) in Farnsworth (2017) to define systems with an inherent cybernetic boundary (where inside is definitively separate from outside) as only those systems having transitive closure for causation, (xCy: read as x causes y), meaning that the state of an object y is  strictly determined by the state of the object x. The transitive causal closure of a system means that its  components form a set A of causally related objects under C such that there is no object in A who’s  state is not caused by an object in A: every part of the system is causally connected to every other.
It is from such systems that a transcendent complex may arise. Finally, taking causation as manifest in mutual information, Bertschinger et al. (2006) derived a quantitative metric of information closure to  operationalise these concepts in systems theory, which has developed into an information-theoretic  method for system identification [113,Bertschinger et al. 2008].






see Autocatalytic Sets.

Information Abstraction




In Farnsworth (2018) I concluded that organism identity is created by the information structure existing at the highest level of causal closure at which the highest level of will-nestedness is identified and that this coincides with the ‘maximally irreducible cause-effect structure’ defined in IIT.


Free Will Machine
The "free will machine" (from Farnsworth 2017) is a kind if cybernetic structure intended to illustrate the requirements for free will as some philosophers define it. The term 'machine' follows the convention in cybernetics of calling information processing devices machines. This one generates predictions of its own state in alternative futures Ft+n (calculated by Turing Machine TM2), having built an internal representation of itself interacting with its environment (this representation is created by TM1). It selects the optimal response from among possible reponses at time t (Rt), using a goal-bases criterion G, within the Finite State Automaton (FSA), in which the goal is internally determined. IPS is the implementation of the selection in the physical system - a translation of information (the optimal future f' into optimal action r' which changes the system's state from Qt to Qt+1 .

Bertschinger, N.; Olbrich, E.; ay, N.; Jost, J. (2006). Information and closure in systems theory. German Workshop on Artificial Life <7, Jena, July 26 - 28, 2006>: Explorations in the complexity of possible life, 9-19.

Bertschinger, N.; Olbrich, E.; ay, N.; Jost, J. (2008). Autonomy: An information theoretic perspective. Bio Systems 2008, 91, 331–45.

Ginsborg, H. (2006). Kant’s biological teleology and its philosophical significance. – In: Bird, G. (ed.), A companion to Kant. Blackwell, pp. 455–470.

Farnsworth, K.D. (2017). Can a Robot Have Free Will? Entropy. 19, 237; doi:10.3390/e19050237

Farnsworth, K.D. (2018). How organisms gained causal independence and how to quantify it. Biology.

Kauffman, S. A. and Clayton, P. (2006). On emergence, agency and organisation. – Phil. Biol. 21: 501–521.

Rosen, R. (1991). Life itself: A comprehensive enquiry into the nature, origin and fabrication of life; Columbia University Press: New York, USA.

Sharov, A.A. (2010). Functional Information: Towards Synthesis of Biosemiotics and Cybernetics. Entropy, 12, 1050–1070.



  • Back to Philosophy Theme