S15,E10B: Everything Determines But Us
S15,E10B: Everything Determines But Us
Series 15: How to Remain Blameless while Calling BS on 'No Free-Will'
1. The Solitary Neuron
In his book, “Determined: The Science of Life Without Free Will”, Sapolsky challenges the ‘free-will-ers’ to show him any neuron that created or triggered a particular decision. Good one right? But why should there be a ‘neuron-zero’ behind every decision? Were I to ask - ‘which of the seven billion transistors on that processor triggered that calculation?’ - it wouldn’t make any less sense than his question. To make a decision or perform a calculation requires an integrated system that can manipulate information be that a person or a computer.
Within his philosophy of causal-intersectionality, Sapolsky permits himself an unlimited number of contributory determining factors, most of which are better understood as context. This is a clue to the trickery his flawed logic is playing on him. While assuming that free-will implies an absence of context and a single point of causation, he doesn’t see that it would be equally valid to apply those arbitrary stipulations to his argument.
Were we to need a solitary neuron with inherent intentionality to demonstrate free-will, then determinists should be able to show that within the series of antecedent events leading up to an action, there is a solitary neuron providing a causal link. Perhaps Sapolsky can direct us to that one. I’ll even grant the leeway that it can be completely determined by its environment and it won’t help him.
Yet didn’t I do something similar? In S15, E3: A Unit of Consciousness I was trying to make the point that there must be a set of minimum requirements for an element of consciousness, be it, some kind of self-observing neuron (unlikely); a circuit of neurons observing each other (more likely) or, as I explored in episodes E5, E6A and E6B (and considered to be most likely), emerging from regions in the brain observing each other.
I was not claiming to make a serious hypothesis about the location of sentience, but was merely stating that to understand consciousness we must determine at what level it emerges. Sandwiched between my discourses on the nature and architecture of sentience, S15, E4: The Unconscious Coupling speculated on the combined roles of the conscious and unconscious processes, in the production of free-will.
If we think of the brain as a machine that manufactures decisions using the raw inputs of energy and information, it’s not difficult to conceptualise control system type explanations for consciousness that are plausible i.e., at the operational scale rather than on relying upon tubules and quantum mechanics.
This is not to say that the construct introduced in E4 is correct, just that it’s a reasonable line of enquiry, liable to be more fruitful than re-engineering society so people can bypass their neurology; I am of course talking about the mechanisms for motivation, reward, pleasure, guilt, shame etc. My case for this being functionally dangerous comes later in this piece.
The Purpose of Consciousness
I suggest that the biggest indication of what can be learned about consciousness, is to think about what it does for us and why. Awareness, enables us to establish a sense of ‘wellness’, and establish our preferences for certain outcomes (see E7). Through a process of sensory fusion it allows us to model our reality and navigate it. Is it odd to suppose that we might act on our preferences and move towards better outcomes in the service of self-interest?
2. Navigating the n-Body Problem
In Astrophysics, the three-body problem concerns the interaction of three orbiting bodies, such that each orbital pathway is influenced by the gravitational attraction of the other two bodies. The resulting mutually confounding system becomes unsolvable in an analytical sense - meaning there is no equation to describe the motion for any given time.
The three-body problem can be incrementally calculated or simulated but overall it’s unpredictable despite being determined. Two chapters in Sapolsky’s book were written to debunk unpredictability and randomness as possible components of free-will.
“Chapter 9: A Primer on Quantum Indeterminacy” is self-evidently titled and it prepares the way for “Chapter 10: Is your free will Random?” There he explains how quantum indeterminacy has been misused as an explanation for free-will by mystics and pseudo-scientists.
As far as I am concerned this is to push on an open door. Instead, consider the possibility that free-will might have evolved to negotiate uncertainty, rather than be built out of it. Hold it for a few seconds. Doesn’t that feel better?
Let me first share some things I found on chaotic systems:
In 2005, while I was a PhD student in Applied Mathematics at the University of Maryland, the legendary Lorenz visited my advisor Eugenia Kalnay in her office in the Department of Atmospheric & Oceanic Science. At some point during his stay, he penned the following on a piece of paper: “Chaos: When the present determines the future, but the approximate present does not approximately determine the future.”
Chaos in an Atmosphere Hanging on a Wall, by Christopher M. Danforth
Forgive the presumption, but I paraphrase the Edward Lorenz quotation to see if I can make it a little more accessible:
The present state of a chaotic system precisely defines its future states but an approximate present state does not give us any indication of an approximate future outcome. In other words, within chaotic systems, small changes in initial conditions produce disproportionately large differences in outcomes. This is about sensitivity to initial conditions. Invoking the three-body example, the information pertaining to the motion of the three bodies is there in the system, just that it can’t be encoded in a mathematical equation. Although the position of those bodies in the near future can be calculated, there is no mathematical expression that can describe the motion for some other arbitrary time in the future (or past for that matter); getting to that future state requires us to calculate all the intermediate states in order.
The Stepwise Cause and Effect Cycle in Chaotic Systems
From initial conditions the state of a system in the immediate future can be calculated. The result of that calculation becomes the new initial conditions from which the next state can be calculated - and so on.Since the environment ‘acting on us’ is comprised of an indefinite number of variables and that many of them are mutually influential, it seems that unpredictability is the norm. But something is unpredictable or uncertain because the information is hidden - but hidden from who? Well, from us, for as long as we don’t have the mathematics to describe it.
I suggest that mathematics is as a corpus of informational languages, the main application of which is to decode the information that exists in our universe. I think this is what Max Tegmark gets wrong in his book, The Mathematical Universe, where he makes the odd and unconvincing suggestion that the universe is actually built out of mathematics.
Call it my ignorance if you like but for similar reasons, the supposedly big problem of ‘fine-tuned physical constants’, is to me an uninteresting non-issue. If, as seems obvious to me, mathematics is a means to decode the information hidden in our universe, wouldn’t any mathematical tools that actually work have to be fine tuned to that universe?
So while generally, the n-body problem is not be solvable (we cannot derive a predictive model that works for any arbitrary point in time), to me it seems more relevant to ask if it is immediately navigable and if so, how? I think the answer might be that humans use consciousness and free-will to mediate their interactions with complex situations.
I might be persuaded that a universe harbouring no sentience would be deterministic, but sentient creatures don’t interact with the environment by ricocheting off solid objects, instead they navigate around them. But what about the claim that having free-will would allow us to inadvertently trigger a ripple of change that would disrupt all our futures? The fact is that systems that are locally linear are not sensitive to initial conditions. Also, there are chaotic systems that have attractors, which means that a system with many possible initial states might map to a much smaller set of outcomes. An intuitive example of this comes in around 727 words.
Channelling the Weather
Humans are pattern-seeking mammals and have developed lots of tools to find connections between things and reveal new information. Those tools now include software modelling and simulation of dynamic systems, and probably the most familiar example to most people, would be weather forecasting.
The tiniest turbulences can have wildly disproportionate impacts on weather systems, or put another way, weather systems are extremely sensitive to initial conditions. The limit of reliable weather prediction is around 14 days and yes, that is because weather and climate systems are chaotic, and the initial conditions for two weeks ahead are all but unknowable.
The way into this problem is to use ensemble simulations. This involves running repeated model simulations with small changes in the initial conditions. It is the level of agreement between these runs that determine the probability of weather events. It is a case of looking for the most likely end state i.e. the one that the simulations appear to be converging on, or what we might call the ‘attractor state’.
However, if the distribution of probable outcomes diverges too much, it means that particular weather system is too sensitive to initial conditions to be reliable, making it impossible to make a confident prediction. This is why the range of prediction is limited to around two weeks, because if these simulations are rolled forward beyond that, the spread of possible initial conditions creates too much uncertainty.
I think it’s free-will that allows us to navigate weather uncertainty, for example by carrying an umbrella and a snow shovel in the car, just in case. To varying degrees of success, uncertainty can be negotiated, on an ad hoc basis. This is also my crude segue into looking at another aspect of chaos, which I will link back to what Lorenz said before explaining what I think is the true relevance all this has to free-will.
Tim Palmer’s Pendulum
Tim Palmer, Oxford University Professor, climatologist and physicist uses the example of an executive decision-making toy, to explain how we can superimpose biases on chaotic systems. This desk novelty consists of a ferrous pendulum that swings above four magnetic bases each one nominally representing ‘a decision’.
First a little explainer on some of the detail that may or may not be obvious.
Overview
The initial conditions are set up by the person arbitrarily lifting and releasing the pendulum. That action sets the pendulum off on a deterministic path to its final stable state, even though that path is chaotic and unpredictable. This might seem contradictory but it’s not.
Upon release, the pendulum accelerates due to gravity towards the point of equilibrium but in this case is also comes under the influence of four magnetic fields which overlays the chaotic motion. The pendulum may oscillate around one of the bases, giving every indication it is about to settle, only to erratically head off to another base entirely - the outcome is determined but unknowable yet for all practical purposes, this is random.
Pendulum Initial Conditions
Let’s think about the starting point for the initiating swing. The pendulum can be started of in a theoretically infinite number of different positions around the base, compounded by factors such as moisture on the fingers or environmental influences. The point is, the set of initial conditions are effectively infinite, not that they are all deliberately selectable at the limit.
Outcomes
In this case the outcome is the final resting place of the pendulum and because four magnets are being used, there are four possibilities. Over many trials the initial conditions are effectively randomised and the stochastic distribution of outcomes would show, given a sufficient number of plays, that the pendulum will settle on each of the four bases 25% of the time. The variability of the initial swing will become less significant as the number of runs increase. However, for any given set of initial conditions there is only one outcome, so this is what is meant when it is said that outcomes are determined despite being unpredictable.
The Point Is?
Here we have an infinite number of initial conditions mapping to four possible positional states, which incidentally, are also chaotic attractors. Although it can be argued that these are deterministic and in accordance with Newtonian mechanics, time-reversible, each outcome has an unlimited number of antecedent pathways. Here is an example where the outcomes (i.e. not pathways) of a chaotic system are paradoxically relatively tolerant of initial conditions. This is what I was referencing approximately 727 words ago while doubting the idea that free-will necessarily disrupts the future of the universe.
Professor Palmer asked his audience at the Perimeter Institute (Ontario, Canada) what would happen if the pendulum was tilted by putting a wedge under the base. His insight was that although the motion would still be chaotic and unpredictable, over many trials the distribution of ‘final location’ would be skewed. The effect of the wedge would be to superimpose a bias which he likens to what climate change does to weather systems. The concept of ‘tipping points’ fits right into this analogy because we can imagine what happens as this increases.
One of the chief difficulties in understanding climate change is that if there are incrementally-increasing superimposed-influences, at some point in time (in the past, present or perhaps for sometime into the future), they will be smaller than natural variability. The thing that is often misunderstood is that even if the noise is bigger than the signal, it doesn’t detract from the fact that there is a signal, and that it can still mean something.
Let me try to clarify with an analogy.
If you have a small leak in your fuel tank but drive many miles, that leak could easily be masked by the natural variability of your tank level. You may only notice if you accurately log your consumption with sufficient resolution or if the leak gets bad enough to be obvious. Regardless of whether you notice it or not it’s still a leak.
I would recommend the lecture Tim Palmer he gave at the Perimeter Institute that can be found below. If you don’t have time for the full thing, do put it on your ‘watch list’ and even if you can only spare two minutes, you can pick it up from 12.44 to get the example directly from him.
Tim Palmer Public Lecture: Climate Change, Chaos, and Inexact Computing
So what does that long diversion have to do with free-will?
Indeterminacy and Brain Architecture
I have written a little bit about the architecture of the brain in previous episodes, having acquired a rudimentary appreciation of the some of the processes, by thinking of it as a control system. Now, the human brain is far from a conventional control system and one way this is true becomes immediately apparent when you look at how the regions of the brain interact.
Let’s start with an abstract hypothetical.
Region A provides an input into region B, both of which feed forward to region C and D; regions C and D feedback to A. I suggest this looks like an n-body problem. Now to create a simple process control system like that would lead to instability or what we call ‘hunting’.
This leads me to think that what we are looking at is really a form of naturally evolved PID1 loop control - albeit an extremely complex embodiment. It’s something I plan to discuss in a different series but the relevant takeaway here is there are sensible comparisons to be made with the relatively simple control system technology humans have already created.
Yet there is a bigger issue to consider related to the stability of control systems. I have already claimed that attempts to bypass the neurological hardware that evolved to provide mechanisms for motivation, reward, pleasure, guilt, shame etc. would be functionally dangerous, making the promise to explain later. That particular ‘later’ is now.
Our thought processes are relatively stable despite having complex specialist brain regions all mutually interacting and confounding each other. Do we suppose that we can consistently over-ride any of these and maintain the brain as a stable system? If an n-body problem suddenly becomes an (n-1)-body problem do we suddenly arrive at a frictionless utopia?
Here is a high level neurological example to bring us back on track.
Hippocampus Interactions Example
The hippocampus has many pathways for both efferent (motor neurons travelling from the nervous system to the body) and afferent (for sensory feedback from the body) telemetry. These also provide connectivity within the brain to provide, inter alia, facility for learning, memory, emotional control and spatial awareness.
The entorhinal cortex is a main interface between the hippocampus and many other regions to form memories, spatial awareness and navigation capability. An area of the hypothalamus called the premammillary region, contributes to instinctual responses relating to sex and aggression, but also influences memory and spatial ability. The septal area provides a modulating function for emotions and regulates reward and pleasure seeking, and is involved with motivation and reinforcement.
The prefrontal cortex is concerned with decision-making and operational memory and via it’s connectivity to the hippocampus much of this is outsourced, plus these pathways provide access to long-term memory storage. This very quickly becomes complicated and we haven’t really started yet. So for example, the anterior cingulate gyrus also assists in emotional regulation, decision making and is hardwired to both the hippocampus and the prefrontal cortex.
The hippocampus is connected to the primitive amygdala (from where primal feelings of fear and pleasure are generated and processed to influence emotionally charged memories) as well as receiving inputs from various cortical areas referred to as the afferent cortex; these assist with memory and sensory fusion. Moving to the brainstem we have the reticular formation that is connected with several regions of the brain including the hippocampus to promote vigilance and focus and is the main command centre for running the automatic functions, such as respiration and heart beat.
Much of this connectivity is bi-directional and most (if not all - I’m not sure) regions either directly or indirectly influence all other regions but does that mean that the outputs are randomly produced? No, and it’s not difficult to see how the brain might be viewed as a chaotic yet nevertheless determined system. It would explain why the outputs from the brain’s neurological processes are unpredictable from the outside. You might conclude from this that free-will doesn’t exist after all, but perhaps there’s another way to look at it, and here’s my attempt at demonstrating the point.
On the inside of this thinking process nothing feels like it is unpredictable. Why might that be? Because we have private information that removes the uncertainty. Even if we might not be consciously aware of all the data that’s available to our unconscious processes it is still a part of who we are and therefore, at some level known to us. We experience the fusion of these processes from the integration of sensory signals to the bubbling up of nebulous intuitions from our unconscious. Consciousness brings with it a sense of self and free-will.
Is Free-Will Directional?
What do I mean by that questioning subtitle? Well, is it plausible that we use free-will to superimpose a directional bias on chaotic systems, allowing us to navigate uncertainty? Could decision-making consist of putting a really big wedge of free-will under the base of the deterministic pendulum? What if the wedge was so large, that the sensitivity to initial conditions and the chaotic component of the system, were often entirely overwhelmed by this superimposed bias (by which I mean a strong preferential drive)?
In addition to being a good reason for consciousness to have evolved, it also allows you to see that my navigation metaphor is better than you thought, so that’s nice.
3. Transitions of Attribution
Sapolsky associates the removal of human attribution with progress and this section outlines my objection and counter to that. This is my summary of this aspect of his argument as I understand it.
My Interpretation of the Sapolsky Argument
Absolutely everything contributes to who we are, what we do as a mechanistic subset of a universe where everything is determined by a causal chain. Whatever we think, decide or do, is determined so what we might think of as cognitively triggered events, but those events are merely dumb links in the causal chain.
We are determined by everything except ourselves, therefore it is inappropriate to blame anyone for anything, since nothing can be their fault.
Similarly it is not appropriate to reward or praise anyone, because that would be undeserved; whatever they have achieved can only be a result of luck and privilege. Scientific methodologies have allowed us to push back against superstition and misattribution.
Because blame has been incorrectly ascribed to classes of people in history, Sapolsky concludes the problem is attributing anything to humans. His concept of progress amounts to the ‘subtracting out’, not just blame, but attribution in general and a casualty of that is meritocracy.We no longer blame lepers for poisoning water wells, witches for communing with the devil or mothers for being schizophrenogenic, in part because the supposed causality has been debunked. Yet this is just not about ‘subtracting out’ but substitution too. Historically, the primitive myths and models about reality have only been properly discredited after being displaced by better and generally bigger explanations.
There is no such thing as an attribution vacuum and the fact that a cause is unknown does not mean it’s not there. This is why we use mathematics to find hidden informational content.
A favourite example of mine is our Sun, which the ancient Egyptians believed was transported by a living god, namely Ra. Ra travelled across the sky in his chariot and at sunset when he would go into the underworld, fight his way through death and then give birth to himself the following dawn. For the Egyptians this was part of a much bigger model of reality through which they were able to explain natural phenomena like sunrise, sunset and the Nile inundation. How could anyone have disproved the existence of Ra in ancient Egypt?
Around 2,600 years later, Claudius Ptolemy’s geocentric model put Earth at the centre and set the observable celestial bodies, including the Sun, in an orbit around it. Humans found themselves to be the focus of everything which perhaps fostered our enduring grandiose sense of purpose. The current heliocentric model was revealed by Nicolaus Copernicus nearly five hundred years ago, but it was suppressed by religious dogma, until it could no longer be contained. In 1992, Pope John Paul II exonerated Galileo Galilei for his support of the Copernican construct, around three-hundred and fifty years too late to stop the Roman inquisition.
Now of course, we have new and spellbinding views of the universe and are beginning to appreciate how insignificant we are in the cosmological scale, with the possible exception of certain fatalistic religions and those who believe the Earth is the only frisbee in the solar system.
Yet old habits die hard. Liberated from being at the mercy of gods, now we can absolved of sin and rendered blameless, by surrendering to something greater than us called determinism. The state alone can make progress if we consent to becoming passengers in our own existence. Only those who repent and attribute-no-more can enter the promised land of determinism.
As is frequently the case with human conceptions of paradise, it’s frictionless, infantilising and anodyne. For this we must sacrifice claims of autonomy, today, everyday and forevermore.
War Is Peace
Freedom Is Slavery
Ignorance Is Strength
“Nineteen-Eighty-Four”, George Orwell
Sapolsky did not stagger down the mountain carrying tablets of deterministic truth but he just might be smelting your next god.
PID (Proportional, Integral, Derivative) controller is a device that modulates control to match a measured variable to a desired set point in real time. The difference between these two values is the error signal so at a given time, the correction is comprised of functions that are representative of the -
- proportionality of the error
- sum of errors
- rate of error change