Ben Goertzel
February 18, 2004
This note presents a novel, system-theoretic explanation of the psychological phenomenon of “free will,” in terms of the dynamics and interactions of different parts of the brain. The theory integrates neuropsychological observations of Libet and Gazzaniga, but it also has a generality that extends beyond human brains to other forms of intelligence, including computational ones. The relationship of free will with the phenomenon of consciousness is also addressed.
Suppose one has a world whose governing dynamic has a high Liapunov exponent (Devaney, 2003), so that a small region of state space at time t is dynamically mapped into a much larger region of state space at size t+s, even if s is small. Then, an intelligent system (let’s call it a “brain” for short, though it may be computational or biological), in order to plan for the future, must create a virtual multiverse inside itself: i.e. at time t it must model several different future states for time t+s, since it doesn’t know which future state will actually occur. It must create a virtual multiverse with branch-points regarding its own external actions, and its own internal events, as well as external events not directly caused by itself. This is what our brains do all the time.
The notion of a “multiverse” here is motivated by quantum mechanics (DeWitt and Seligman, 1974). However, the theory I am proposing here is not a quantum theory of consciousness; it is compatible with both quantum and classical physics. What I mean by a multiverse is a model of reality like the one explored by Borges in his famous tale The Garden of Forking Paths (see Borges, 1999). Borges portrayed the world as consisting of pathways defining series of events, in which each pathway eventually reaches a decision-point at which it forks out into more than one future pathway. Borges’ “paths” are the “branches” of the mathematical “tree structures” used to model multiverses; and his decision points are the nodes or “branch-points” of the trees. Actual reality is then considered as a single “universe” which is a single series of events defined by following one series of branching-choices through the mathematical tree. The many-universes interpretation of quantum physics posits that the multiverse is physically real, even though we as individuals only see one universe; and that an act of quantum measurement consist of a choice of direction at a branching point in the multiverse tree. On the other hand, what I am hypothesizing here is that we perceive a psychologically real multiverse – independently of whether there is a real physical multiverse or not – and that free will has something to do (details to follow shortly) with the choice of directions at branch-points in this psychological multiverse.
We know that the cognitive portions of brains do not directly experience the external universe; they only experience their own models of the external universe. This is demonstrated by many experiments regarding perceptual illusions, for example (Maturana and Varela, 1992). What this means is that, even if we should happen to live in a strictly deterministic universe[1], we subjectively live in a multiverse in which several different possible branches are subjectively real at any given time. But most of these branches are very short-lived: they exist only conjecturally while we wait for the next percepts which will tell us which of the branches is actualized.
Furthermore, brains largely experience themselves only via their models of themselves. Brains, being complex systems, are hard to predict even for themselves, and so one part of a brain often must use a virtual multiverse to model another part.
When a brain triggers a real-world action, this action occurs in the external universe, and then registers internally in the virtual multiverse which models the external universe. The brain is then aware of a process of “collapse” wherein the multiple branches of the virtual multiverse collapse to a single branch. Furthermore, this collapsing process occurs rapidly, within the same subjectively experienced moment as the actual event in the physical universe. Note that a subjectively experienced moment is not instantaneous.
Similarly, when a part of a brain carries out an action, and another part of the intelligent system is modeling this first part using a virtual multiverse, then the action in the first part corresponds with a collapse to a single branch in the virtual multiverse contained in the second part.
The special feeling of “free will” that we experience consists primarily of the subjectively-simultaneous consciousness of
or else the simultaneous consciousness of
The subjective simultaneity is only present when the two things occur at almost the same physical time, which generally occurs only when the event in question is either internal, or else an external event that’s directly triggered by the brain itself.
Libet (2000) has done experiments showing that, in many cases, the “decision” to carry out an action occurs after the neural signals directly triggering the action have already occurred. This observation fits in perfectly with the virtual multiverse theory. Note that this time interval is sufficiently short that the action and the decision occur within the same subjectively experienced moment. In fact, Libet’s results, though often presented as counterintuitive, are explained naturally by the current theory – it’s the opposite result, that perceived-virtual-multiverse-collapses occurred after the corresponding actions, that would be more problematic for the current theory.
Dennett (2003) analyzes Libet’s results by positing that free will is a distributed experience which occurs over an expanse of time (the experienced moment) and a number of different brain systems, and that there is nothing paradoxical about the part of this experience labeled “decision” occurring minutely before the part of this experience labeled “action trigger.” I agree with Dennett’s general observations – and with most of his comments about free will – but I am aiming to achieve a greater level of precision in my analysis of the phenomenon.
For example, suppose I am trying to decide whether to kiss my beautiful neighbor. One part of my brain is involved in a dynamic which will actually determine whether I kiss her or not. Another part of my brain is modeling that first part, and doesn’t know what’s going to happen. A virtual multiverse occurs in this second part of the brain, one branch in which I kiss her, the other in which I don’t. Finally, the first part comes to a conclusion; and the second part collapses its virtual multiverse model almost instantly thereafter.
The brain uses these virtual multiverse models to plan for multiple contingencies, so that it is prepared in advance, no matter what may happen. In the case that one part of the brain is modeling another part of the brain, sometimes the model produced by the second part may affect the actions taken by the first part. For instance, the part (call it B) modeling the action of kissing my neighbor may come to the conclusion that the branch in which I carry out the action is a bad one. This may affect the part (call it A) actually determining whether to carry out the kiss, causing the kiss not to occur. The dynamic in A which causes the kiss not to occur, is then reflected in B as a collapse in its virtual multiverse model of A.
Now, suppose that the timing of these two causal effects (from B to A and from A to B) is different. Suppose that the effect of B on A (of the model on the action) takes a while to happen (spanning several subjective moments), whereas the effect of A and B (of the action on the model) is nearly instantaneous (occurring within a single subjective moment). Then, another part of the brain, C, may record the fact that a collapse to definiteness in B’s virtual multiverse model of A, preceded an action in A. On the other hand, the other direction of causality, in which the action in A caused a collapse in B’s model of A, may be so fast that no other part of the brain notices that this was anything but simultaneous. In this case, various parts of the brain may gather the mistaken impression that virtual multiverse collapse causes actions; when in fact it’s the other way around. This, I conjecture, is the origin of our mistaken impression that we make “decisions” that cause our actions.
The “illusion” of free will, therefore, consists largely of a mistaken impression gathered by some parts of the brain about the ordering of events in other parts of the brain. It consists of a confusion between two different roles played by virtual multiverse models:
Because in the former, multiple-subjective-moment case, virtual multiverse collapse precedes action-determination, the brain mistakenly infers that in the latter, single-subjective-moment case, virtual multiverse collapse also precedes action-determination. But in fact, in the latter case virtual multiverse collapse follows action-determination.
However, it is not an illusion or confusion that virtual multiverse modeling has an impact on actions taken in the brain. This kind of modeling is clearly a very valuable part of brain dynamics, due to the complex and hard-to-predict nature of the brain and world. Virtual multiverse modeling is necessary due to practical indeterminism within and outside the brain, which exists whether or not fundamental indeterminism does. It is necessary because internal and external events are often indeterministic from the subjective perspective of particular, useful parts of the brain. Furthermore, and critically, the brain as a whole is often indeterministic from its own perspective.
Another side of free will is the “confabulative” aspect emphasized by Michael Gazzaniga in his discussions of his famous split-brain experiments. These experiments demonstrate that, even when there is a clear external cause of a human taking some action, it is possible for the human to sincerely and thoroughly believe that the cause was some completely internal decision that they took. The left hemisphere of a split brain has no experience of stimuli delivered exclusively to the right hemisphere (e.g. through the left eye). However, the left hemisphere has such a strong motivation to create explanations that it will make up “free will stories” corresponding to behaviors initiated by the isolated right hemisphere. For example, in one experiment, a split brain subject's left eye received a command to stand. The person stood – and then, when asked why she stood up, she responded (using the language center of her left hemisphere) that she wanted a soda. In another experiment, when the left and right hemispheres were each asked to pick an appropriate picture to accord with an image flashed only to that hemisphere, the left selected a chicken to match the chicken claw in the picture it saw, while the right hemisphere correctly chose a shovel to remove the snow it saw. When asked why the person chose those images, he replied that the claw was for the chicken, and the shovel was to clean out the shed (Gazzaniga, 1989).
Confabulation means that, when a certain branch in the virtual multiverse has been chosen, the brain looks for reasons why it was chosen. If no immediate reasons are available, it will use inference to create reasons. Often these inferences will be accurate; sometimes they will be erroneous. Split brain surgery creates a situation in which erroneous inference of this nature are much more common than usual. It happens that in humans this explanation-generating inference tends to take place in the left brain hemisphere; but the same post-facto explanation-generating dynamic may be expected to exist in nonhuman intelligences as well, regardless of whether their brains display any hemispheric dichotomy.
Confabulation adds a third aspect to virtual multiverse dynamics: not only do virtual multiverse inferences/simulations affect actions, and actions cause updating of virtual multiverse simulations; but also, reasoning about actions causes inferred stories to be attached to the memories of virtual-multiverse collapses.
Finally, the present theory of free will also partially addresses the phenomenon
of consciousness. Some aspects of
consciousness can be understood by thinking about the virtual multiverse models
that parts of the brain construct, in order to model the brain as a whole. These virtual multiverse models are used to
help guide the dynamics of the whole brain (on a slow time scale), and they are
also continually updated to reflect the actual dynamics of the brain (on a
faster time scale, occurring within a single subjective moment). The feeling of consciousness is in part the
feeling of events in the whole brain being rapidly reflected in the changes in
the virtual multiverse models maintained in parts of the brain … and these
changes then causing further virtual-multiverse-model changes which then feed
back to change the state of the whole brain again … etc. The conscious feeling of the flow of time is
actually a feeling of continual ongoing branch-selection in the virtual
multiverse model of the whole brain – the feeling of briefly-explored possible
futures being left by the wayside as the actualized futures are registered in
the model.
Dennett (1992) analyzed human consciousness as a serial computer running as a virtual machine on top of a parallel computer (the “parallel computer” being the unconscious, which comprises the majority of brain function). However, I don’t think this is quite right. Rather, I think human consciousness has to do with the feedback between virtual multiverse modeler software (embodied in various parts of the brain) and massively parallel software (the rest of the brain). The virtual universe modeler software is not exactly a serial computation process, it may well explore multiple branches in parallel.
The virtual-multiverse theory of free will does not explicitly solve the “hard problem of consciousness” (Chalmers, 1997), the relationship between subjective awareness (“qualia”) and physical phenomena. However, it does fit in naturally with a particular hypothetical solution to the hard problem. Suppose one accepts, as a solution to the hard problem, the postulate that a quale occurs when a system comes to display a pattern that it did not display a moment before; and the more prominent patterns correspond to the more intense qualia.[2] Then, it follows from the present theory of free will that intense qualia will tend to be correlated with significant activity in the whole-brain virtual multiverse modeler. This provides an explanation for the oft-perceived correlation between consciousness and free will (free will also often being associated with significant activity in the whole-brain virtual multiverse modeler).
What I have proposed here is a conceptual model of free will in terms of virtual multiverse modeling, but it also leads to some specific empirical predictions. Study of the human brain, as brain imaging improves, should allow us to localize the brain’s multiverse modeling faculties (assuming these exist, as I hypothesize), and then to study whether the dynamics of interaction between these faculties and the rest of the brain are indeed as I have hypothesized. Regarding artificial intelligence, the hypothesis made is that if an AI program is created with a virtual-multiverse-modeling faculty that is embedded into its overall dynamic process in a manner roughly similar to how this embedding occurs in the human mind/brain, then the AI will describe its decision-making experiences in roughly the same way that humans describe their experience of free will.
· Borges, Jorge Luis (1999). Collected Fictions. Viking.
· Chalmers, David (1997). The Conscious Mind. Oxford University Press.
· Dennett, Daniel (1992). Consciousness Explained. Back Bay Books.
· Dennett, Daniel (2003). Freedom Evolves. Viking.
· Devaney, Robert (2003). An Introduction to Chaotic Dynamical Systems. Westview Press.
· Dewitt, Bryce and C. Seligman (1974). The Many-Worlds Interpretation of Quantum Mechanics. Princeton University Press.
· Gazzaniga, Michael (1989). "Organization of the Human Brain," Science, Sept., pp. 947-956
· Libet, B., A. Freeman and K. Sutherland (2000). The Volitional Brain: Towards a Neuroscience of Free Will. Imprint Academic.
· Maturana, Humberto and Francisco Varela (1992). The Tree of Knowledge. Shambhala.
· Peirce, Charles S. (1982). Collected Works Volume 5. Indiana University Press.
[1] In fact, I do not believe we live in a purely deterministic universe; but I think this issue should be separated from analysis of the phenomenon of free will
[2] In terms of the philosophy of Charles Peirce (1982), this postulate suggests that Firsts correspond to newly minted instances of Thirds