Can we talk scientifically about free will?

(edited and updated – 3 May 2019)

For some, the scientific way of thinking is both challenging and attractive.  Thinking scientifically leads to an introduction to, and sometimes membership in a unique community, who at their best are curious, critical, creative, and receptive to new and mind-boggling ideas, anchored in objective (reproducible) observations whose implications can be rigorously considered (1).  

What I particularly love about science is its communal aspect, within which the novice can point to a new observation or logical limitation, and force the Nobel laureate (assuming that they remain cognitively nimble, ego-flexible, and interested in listening) to rethink and revise there positions. Add to that the amazing phenomena that the scientific enterprise has revealed to us, the apparent age and size of the universe, the underlying unity, and remarkable diversity of life, the mind-bending behavior of matter-energy at the quantum level, and the apparent bending of space-time.  Yet, and not withstanding the power of the scientific approach, there are many essential topics that simply cannot be studied scientifically, and even more in which a range of practical constraints seriously limit our ability to come to meaningful conclusions.  

Perhaps acknowledging the limits of science is nowhere more important than in the scientific study of consciousness and self-consciousness.  While we can confidently dismiss various speculations (often from disillusioned and displaced physicists) that all matter is “conscious” (2), or mystical speculations on the roles of  super-natural forces (spirits and such), we need to recognize explicitly why studying consciousness and self-consciousness remains an extremely difficult and problematic area of research.  One aspect is that various scientific-sounding pronouncements on the impossibility or illusory nature of free will have far ranging and largely pernicious if not down right toxic social and personal  implications. Denying the possibility of free will implies that people are not responsible for their actions – and so cannot reasonably be held accountable.  In a broader sense, such a view can be seen as justifying treating people as disposable machines, to be sacrificed for some ideological or religious faith (3).  It directly contradicts the founding presumptions and aspirations behind the enterprise that is the United States of America, as articulated by Thomas Jefferson, a fragile bulwark against sacrificing individuals on the alter of often pseudoscientific or half-baked ideas.

So the critical question is, is there a compelling reason to take pronouncements such as those that deny the reality of free will, seriously?   I think not.  I would assume that all “normal” human beings come to feel that there is someone (them) listening to various aspects of neural activity and that they (the listener) can in turn decide (or at the very least influence) what happens next, how they behave, what they think and how they feel.  All of which is to say that there is an undeniable (self-evident) reality associated with self-consciousness, as well as the feeling of (at least partial) control. 

This is not to imply that humans (and other animals) are totally in control of their thoughts and actions, completely “free” – obviously not.  First, one’s life history and the details of a situation can dramatically impact thoughts and behaviors, and much of that is based on luck, a range of hereditary factors, our experiences (both long and short term) that combine to influence our response to a particular situation – recognition of which is critical for developing empathy for ourselves and others (see The radical moral implications of luck in human life).  At the same time how we (our brain) experiences and interprets what our brain (also us) is “saying” to itself is based on genetically and developmentally shaped neural circuitry and signaling systems that influence the activities of complex ensembles of interconnected cellular systems – it is not neurons firing in deterministic patterns, since at the cellular level there are multiple stochastic processes that influence the behaviors of neural networks. There is noise (spontaneous activity) that impacts patterns of neuronal signaling, as well as stochastic processes, such as the timing of synaptic vesicle fusion events, the cellular impacts of diffusing molecules, the monoallelic expression of genes (Deng et al., 2014; Zakharova et al., 2009) and various feedback networks that can lead to subtle and likely functional differences between apparently identical cells of what appear to be the “same” type (for the implications of stochastic, single cell processes see: Biology education in the light of single cell/molecule studies).

So let us consider what it would take to make a fully deterministic model of the brain, without considering for the moment the challenges associated with incorporating the effects of molecular and cellular level noise. First there is the inherent difficulty (practical impossibility) of fully characterizing the properties of the living human brain, with its ~100,000,000,000 neurons, making ~7,000,000,000,000,000 synapses with one another, and interacting in various ways with ~100,000,000,000 glia that include non-neuronal astrocytes, oligodendrocytes, and immune system microglia (von Bartheld et al., 2016). These considerations ignore the recently discovered effects of the rest of the body (and its microbiome) on the brain (see Mayer et al., 2014; Smith, 2015).

Then there is the fact that measuring a system changes a system. In a manner analogous to the Heisenberg uncertainty principle, measuring aspects of neuronal function (or glial-neural interactions) will necessarily involve perturbations to the examined cells – recent studies have used a range of light emitting reporters to follow various aspects of neuronal activity (see Lin and Schnitzer, 2016), but these reporters also perturb the system, if only through heating effects associated with absorbing and emitting light. Or if they, for example, serve to report the levels of intracellular calcium ions, involved in a range of cellular behaviors, they will necessarily influence calcium ion concentrations, etc. Such high resolution analyses, orders of magnitude higher than functional MRI (fMRI) studies  would likely kill or cripple the person measured. The more accurate the measurement, the more perturbed, and the more altered future behaviors can be expected to be and the less accurate our model of the functioning brain will be.

There is, however, another more practical question to consider, namely are current neurobiological methods adequate for revealing how the brain works.  This point has been made in a particularly interesting way by Jonas & Kording (2017) in their paper “Could a neuroscientist understand a microprocessor?” – their analysis indicates the answer is “probably not”, even though such a processor represents a completely deterministic system. 

If it is not possible to predict the system, then any discussion of free will or determinism is mute – unknowable and in an important scientific sense uninteresting. In a Popperian way (only the ability to predict and falsify interesting predictions makes, at the end of the day, something scientifically useful.  

I have little intelligent to say about artificial intelligence, since free will and intelligence are rather different things. While it is clearly possible to build a computer system (hardware and software) that can beat people at complex games such as chess (Kasparov, 2010; see AlphaZero) and GO (Silver et al., 2016), it remains unclear whether a computer can “want” to play chess or go in the same way as a human being does.  We can even consider the value of evolving free will, as a way to confuse our enemies and seduce love interests or non-sexual social contacts. Brembs  (2010) presents an interesting paper on the evolutionary value of free will in lower organisms (invertebrates).

What seems clear to me (and considered before: The pernicious effects of disrespecting the constraints of science) is that the damage, social, emotional, and political, associated with claiming to have come to an “scientifically established” conclusion on topics that are demonstrably beyond the scope of scientific resolution, conclusions that make a completely knowable and strictly deterministic universe impossible to attain) should be explained and understood to both the general public and stressed on and by the scientific and educational community.  They could be seen as a form of scientific malpractice that should be, quite rightly, dismissed out of hand. Rather than become the focus of academic or public debate, they are best ignored and those who promulgate them, often out of careerist motivations (or just arrogance) should be pitied, rather than being promoted as public intellectuals to be taken seriously.A note on images: Parts of the header image are modified from images created by Tom Edwards (of WallyWare fame) and used by permission. The “Becky O” Bad Mom card by Roz Chast is used by permission.  Thanks to Michael Stowell for pointing out the work of Jonas and Kording.  Also it turns out that physicist Sabine Hossenfelder has recently had something to say on the subject.  Minor updates and the re-insertion of figures – 26 October 2020.

Footnotes 

1. We won’t consider them at their worst, suffice it to say, they can embrace all that is wrong with humanity, leading to a range of atrocities.

3. The universe may be conscious, say prominent scientists

4. A common topic of the philosopher John Gray: Believing in Reason is Childish

Literature cited:

Brembs, B. (2010). Towards a scientific concept of free will as a biological trait: spontaneous actions and decision-making in invertebrates. Proceedings of the Royal Society of London B: Biological Sciences, rspb20102325.

Deng, Q., Ramsköld, D., Reinius, B. and Sandberg, R. (2014). Single-cell RNA-seq reveals dynamic, random monoallelic gene expression in mammalian cells. Science 343, 193-196.

Kasparov, G. (2010). The chess master and the computer. The New York Review of Books 57, 16-19.

Lin, M. Z. and Schnitzer, M. J. (2016). Genetically encoded indicators of neuronal activity. Nature neuroscience 19, 1142.

Jonas, E., & Kording, K. P. (2017). Could a neuroscientist understand a microprocessor?. PLoS computational biology, 13, e1005268.

Mayer, E. A., Knight, R., Mazmanian, S. K., Cryan, J. F., & Tillisch, K. (2014). Gut microbes and the brain: paradigm shift in neuroscience. Journal of Neuroscience, 34, 15490-15496.

Silver et al. (2016). Mastering the game of Go with deep neural networks and tree search. nature 529, 484.

Smith, P. A. (2015). The tantalizing links between gut microbes and the brain. Nature News, 526, 312.

von Bartheld, C. S., Bahney, J. and Herculano‐Houzel, S. (2016). The search for true numbers of neurons and glial cells in the human brain: a review of 150 years of cell counting. Journal of Comparative Neurology 524, 3865-3895.

Zakharova, I. S., Shevchenko, A. I. and Zakian, S. M. (2009). Monoallelic gene expression in mammals. Chromosoma 118, 279-290.