INVESTIGATIONS

THE NATURE OF AUTONOMOUS AGENTS

AND THE WORLDS THEY MUTUALLY CREATE

STUART A. KAUFFMAN

SEPTEMBER 13, 1996

EPILOGUE

A TENTATIVE PHYSICAL HYPOTHESIS CONCERNING CONSCIOUSNESS

E.1.0) The obvious hesitations: Having studied philosophy of mind years ago at Oxford, I am fully aware of how little has beensaid that is sensible on the subject of consciousness. Here is not the place to discuss the well-known and less well-knowndiscussions of this deepest of puzzles. "We seek to understand the visual system and its role in consciousness, not yet theblueness of blue," said Francis Crick recently. But of course, ultimately, the blueness of blue is just the question, as well as whythe appreciation of the sky's normal color seems to occur in higher vertebrates, perhaps other living things, but, we suppose, notin nonliving things.

E.1.1) Given that so little helpful and so much useless has been written on the subject of consciousness, the following tentativehypothesis can hardly do worse.

E.1.2) I am hopeful that the hypothesis is at least sensible, or can be made so, and will have testable consequences.

E.2.0) Several features of the discussion in the first seven lectures above drive in the direction of the following hypothesis:

Consciousness is associated with the fine grained "decision making" leading to alternative behaviors withinmolecular autonomous agents that spans the quantum and classical realms and is identical to the persistentpropagation of percolating webbed loops of quantum coherence that simultaneously are persistently losing coherencehence going classical. The passage to classical behavior is identical to "mind" influencing "matter." Consciousnessis the inner experience of the agent of this percolating web of persistent coherence decohering into classicalbehavior.

E.3.0) The Arguments:

E.3.1) In section 6.8.1 in discussing the implications of the expanding adjacent possible along non-ergodic trajectories in the 6Ndimensional classical phase space, I was driven to say that for this to occur, then along the trajectory, successive equal size smallboxes in the 6N space must have arrows emerging to an ever greater number of adjacent boxes. These adjacent boxes constitutethe adjacent possible from the box in question. In turn, for a deterministic system, this implies that the initial box is broken into atleast as many subvolumes, each with states flowing to a given adjacent box, as there are adjacent boxes. Thus, growth of theadjacent possible implies finer partitioning of successive "source boxes" along the non ergodic trajectory. But the finer partitioningis equivalent to symmetry breakings within that source box. So internal symmetries must be progressively broken allowing flowfrom within that source box to proceed in different macroscopic directions given small differences in state within the source box.

E.3.2) In terms of molecular Autonomous Agents, the partitioning of a source box in the 6N space means that a box representingone or many agents plus their environment has the property that small changes within the state leads to macroscopic differentdirections of flow.

E.3.3) From the perspective of the Agent, this is equivalent to the emergence by exaptations of finer grained "thresholds" foralternative behaviors. This picture is consistent with one in which Agents in communities tune total dynamical trajectories withinthe community to be slightly convergent, hence each agent coarse grains its world with maximal discrimination consistent withacting without trembling hands in a noisy environment partially due to other agent's actions.

E.3.4) It is a physical fact that the finest grained subvolumes in the 6N dimensional space between which different macroscopicflows can be launched, or "decided" must be set by the Heisenberg Uncertainty Principle. Thus, "decision making" amongalternatives by Agents, hence something about the dimensionality of the adjacent possible is bounded by the HeisenbergUncertainty principle.

E.3.5) In principle, autonomous agents might be entirely classical, as far as the discussion of the first seven lectures is concerned.

E.3.6) But is an observed fact that all free living cells, the minimal autonomous agents we know, are in fact molecular systemswith on the order of several trillion molecules, enclosed in bilipid membranes, and operate at a scale where quantum effects do, infact, have macroscopic classical consequences. Thus, rhodopsin in the rods of the eye respond to single quantum events due tophoton absorption. So does chlorophyll. Yet the consequences within the cell decohere to classical behavior. The bacteriumswims in a given direction for dinner, not in a superposition of all possible directions.

E.3.7) The fact that free living cells do span quantum to classical behavior seems to be a powerful fact and clue. The evolutionarypressure to maximize sensitivity and the capacity to carry out alternative behaviors has, by exaptation and natural selection, led tothis spanning, one presumes. Simultaneously, this is precisely part of the biosphere's expansion of its total adjacent possible.

E.4.0) Consider a non-agent system, say of macroscopic complex, classical inanimate objects such as rocks, and the interactionwith a small quantum coherent system. Then DeWitt would argue that this interaction "collapses" the wave function of thequantum system, leading to a choice of Universes. In the Zurek interpretation, the interaction of the quantum and complexclassical system leads to loss of phase information from the quantum system that cannot be reassembled. This decoherence ISthe passage to classical behavior. In the Bohm Hiley interpretation, this interaction leads to loss of active information into inactivechannels which become so twisted that phase information cannot be reassembled, while the wave function persistently expands inthe active channel in just such a way as to balance the loss of phase information in the inactive channels. So, on the B.H.interpretation, quantum behavior persistently propagates somewhere.

E.4.1) On the B-H interpretation, the propagating active channel might propagate in a variety of directions. It need not, butmight, loop back on itself. We consider these two possibilities next.

E.4.1.1) If the propagating wave in the active channel never curves back on itself, then the regions where the wave function isdecohering propagate to indefinitely new regions.

E.4.1.2) If the propagating wave in the active channel curves back on itself in richly webbed loops, then the active phasecoherent information can persistently propagate and persist among these loops, yet persistently also flow along inactive channelssuch that it is also persistently losing coherence along those inactive channels.

E.4.1.3) The topology of the webbed connections allowing such persistent coherent propagation and loss of phase informationinto inactive channels might fall apart into tiny loops isolated from one another, or might be a vast linked web of loops of activechannels. In the latter case, the active information would percolate across the entire system. In the former case it would not.These differences appear to genuinely different physical possibilities.

E.4.1.4) Only some physical systems will have the property of possessing vast linked webs of loops of active channels by whichpersistently propagating active information is passing, while simultaneously, inactive channels are persistently losing phaseinformation hence decohering.

E.4.1.5) In so far as I postulate that such persistent propagation of active information around percolating loops whiledecoherence is also persistently occurring constitutes consciousness, only certain types of physical systems have thisproperty.

E.5.0) Decoherence, the passage from quantum to classical behavior, leaves the decoherent physical system in a particular state,for example with a particular orientation in 3 space with respect to other classical systems. The creation of a particular classicalstate appears to be the analogue of macroscopic work. In addition, it is the creation of a constraint by which the decoherence ofother quantum processes can become oriented. Hence the work can propagate. (I am indebted to Phil Anderson for pointing outthat, even at equilibrium, a quantum system can break symmetry and assume a definite direction.)

E.5.1) Molecular autonomous Agents achieve closure in at least two spaces - catalytic task space, and work-task space suchthat all the components have their formation catalyzed, and all the work tasks are linked in complex webbed causal cycles suchthat all the constraints are constructed and all the context dependent work-tasks are accomplished.

E.5.2) Thus, molecular autonomous Agents are precisely the kind of organization that has closed classical causal loops in richwebs. Therefore, if any of the events along such causal loops are quantum events - photons hitting rhodopsin or chlorophyll,molecular conformational rearrangements in membranes and membrane buried proteins, etc., then molecular autonomous agentshitting against the Heisenberg uncertainty limit in their "decision" subvolumes" are precisely the kinds of physical systems whichshould typically be fine candidates to have closed linked web loops able to propagate quantum coherent active information.

E.5.3) But notice that, merely by virtue of having closed causal pathways in the classical sense, autonomous agents need notpropagate coherent quantum flow in active channels. It might be the case that each "bubble" of a quantum event decoheres intofully classical behavior along the webbed causal chain before the propagating work encounters any further quantum event"bubble."

E.5.4) Conversely, subtle changes might allow a percolating web of looped quantum "bubbles" to coalesce into linked webbedloops along which quantum coherent active information could propagate and simultaneously decohere along side inactivechannels. Further, subtle changes could couple and uncouple such quantum loops into large of small islands, isolated from oneanother by classical regions of the system.

E.5.5) The probability that quantum coherent active information persistently propagates seems likely to be related to the richnessof looped web. If the system has only tiny islands isolated by classical regions, perhaps the persistent coherent propagationquickly dwindles due to rapid loss into inactive channels. The entire small loop rapidly decoheres into classical behavior. In arichly webbed structure, the active information can perhaps more persistently propagate without entirely dissipating irreversiblyinto decoherence classical behavior.

E.6.0) Given the above, the postulate is that persistently propagating quantum coherent behavior along some loops, whiledecoherence is occurring along side channels IS, via the decoherence into specific states, "mind" acting on matter.

E.7.0) Given the above, the persistent propagating coherent quantum information around webbed loops together with itspersistent decoherence IS consciousness as internally experienced by the Agent.

E.7.1) Then persistent consciousness depends critically on a sensitive balance that allows propagating coherent information topersist and regenerate around webbed loops as fast or faster than irreversible loss into decoherence via inactive channels.

E.7.2) The above suggests that subtle changes in the physical properties of the molecular autonomous agent can render itconscious or "sporadically" conscious in local regions where a quantum bubble arises and decoheres.

E.8.0) It is interesting to note the similarity between Bohr's "unanalyzable whole," the Bohm Hiley interpretation in which thequantum potential is not prefixed in the relations between the quantum particles, but lies in the full configuration space of the Nparticles, and the context dependent character both of exaptations as emergent non-algorithmic discoveries of functionalities, andthe stumbling upon task solutions in our creative moments - the tractor engine block's rigidity AS the solution to how to build achassis in which the block itself becomes the chassis.

E.8.1) Possible potential implications of non-locality for such problem solving.

E.9.0) Virtues of the above hypothesis:

i. It is derived from the general assumption that Agents, to enhance sensitivity and work and adjacent possible work space, willbe driven up against Heisenberg Uncertainty Principle.

ii. Thus, such agents should span quantum and classical behaviors in their decision making.

iii. Real free living cells do in fact span quantum and classical behaviors in their decision making. This is very suggestive evidencefor the hypothesis.

iv. Only certain systems can propagate coherent quantum active information while at the same time decohering, thereby creatinga definite state of matter.

v. Thus consciousness would be expected to be most likely to arise in complex molecular autonomous agents as they break

symmetries allowing finer grained flow into different directions in the classical adjacent possible.

vi. The creation of a definite state of matter is, I hope, consistently interpretable as "mind" acting on "matter" as the agent"chooses" where to go in the classical adjacent possible via propagating classical work achieved by linked decoherence patterns.

vii. If the non-ergodic flow of the Universe tends to enhance the dimensionality of the adjacent possible, and creates autonomousagents readily, then the evolution of consciousness should not be rare.

viii. These hypotheses should ultimately have testable consequences, for example in waking and anesthetized subjects. In thelatter anesthetized brain, quantum coherent loops should not propagate, but might either break into small islands, or decoherefaster than in the waking brain. There should be many testable consequences. Among them, the degree of percolation of quantumcoherences among loops might be a measure of degree of consciousness.

ix. It should be stressed that there are similarities between the hypothesis advocated here and that suggested over the pastseveral years by Roger Penrose. Penrose suggests that some new physics associated with the emergence of classicity in neuraltissue, perhaps associated with quantum gravity, may play a role in consciousness. Penrose bases his arguments on the capacityof humans to discover theorems which cannot be formally derived from their axioms. I find his arguments congenial, butindependent of the lines of evidence I adduce above suggesting that autonomous agents may optimize their capacity to makediscriminations by spanning the quantum and classical realms, and do indeed do so.

x. Phil Anderson has expressed very strong skepticism that quantum coherence will be found at body temperature in neuraltissues. His clear concern is that thermal fluctuations will destroy any quantum coherence. While the point has clear merit, it maynot be utterly convincing. In a system such as that envisioned in which work constructing constraints allows the ordered releaseof energy which itself is propagating work, the current constraints must represent temporarily frozen degrees of freedom.Temperature is a macroscopic property of a system, averaged over a volume. It remains very much to be seen whether themicroscopic occurrence of frozen degrees of freedom, perhaps forming percolating webs as in the analysis of Boolean networksat the edge of chaos, may not have a local effective temperature which is low enough to support quantum coherent events. A fewvery tentative lines of evidence support this possibility: 1) Many workers think the molecules, including water, in a cell are all veryhighly organized--perhaps consistent with ever shifting percolating webs of frozen degrees of freedom. 2) Genetic regulatorynetworks, as noted above based on biases in canalizing Boolean functions used among regulated genes, appear consistent with aparallel processing genetic regulatory network in the ordered regime, near the edge of chaos. But cells are in fact carrying outpropagating work attaining task closure to construct themselves and reproduce. If evolution has achieved an ordered statedynamically, with percolating frozen components, and this indeed is the optimal way to carry out propagating work-signaling-deciding by parallel processing molecular autonomous agents, then a percolating web of frozen degrees of freedom is expectedon theoretical grounds. Thus, again, it remains an open question whether quantum coherent properties can arise in suchpercolating webs of frozen degrees of freedom - or elsewhere in cells.

xi. Thus, the hypothesis that quantum events are important in consciousness is worthy of serious consideration.

xii. A possible experimental approach would use Drosophila genetics to select for mutants which are anesthetized by ether orCO2 at successively lower concentrations, and other mutants which require successively higher concentrations, then seek thegenes and proteins involved in this increased or decreased sensitivity to loss of consciousness. The molecules found might beclues to the cellular molecular mechanisms associated with consciousness--and quantum coherent events.

xiii. Other approaches might include use of "squeezed light" or narrow spectra of wavelengths and observation of return lightwavelengths from stimulated samples.

xiv. In summary, the hypothesis that the sustained simultaneous propagation of quantum coherence and its persistent lapse intodecoherence IS consciousness seems not obviously false, and worthy of further consideration.

CAVEATS

These notes began with the reminder that "Investigations" is intended as serious proto-science, not yet as science. My aim hasbeen to try to formulate questions and a tentative framework in which to answer them. Even if some or much of the above provesto be sensible, only the earliest steps have been taken. I do not mistake proto-science for science, and ask that my readers bearthese caveats in mind.