Towards a science of free will

It would be better to follow the myth about the gods than to be a slave 

to the fate of which the physicists speak, Epicurus, 341 BC

Humans have sought to escape the fear that the universe is deterministic and the possibility that we do not have free will for thousands of years. Arguably, at the root of this fear lie "folk" intuitions that our thoughts and conscious awareness are central to many if not all our (non-reflexive) actions and decisions. Yet, over the past few hundred years many human-centered ideas about our universe have been abandoned: the earth is not the center of the universe (Copernicus), biological evolution does not require design (Darwin) and computing does not require understanding (Turing).


And now there is emerging evidence from psychology and neuroscience that our conscious thoughts might not be at the center of our decisions or actions - and that our intuitive sense of free will is misleading (or an "illusion").

Transforming free will language into operationalizable terms

Over the past several decades numerous theories of free will have been proposed rooted in philosophy, philosophy, biology and physics. Many, if not most of these theories, are challenging if not impossible to test. For example empirical theories that suggest free will is an effect of quantum uncertainty propagating to the level of behaviour are very difficult to test as we must follow the chain of causality from the atomic to behaviour scales.  Many others rely on ordinary language, counterfactuals and intuitions to build complex arguments, for example, that free will is a human faculty of being "reasons responsive". Many such theories often beg the question, i.e. they assume the conclusion (e.g. only humans have free will) or rely on terms such as "intention" or "moral responsibility" difficult to operationalize and to test in a lab.

Yet despite these concerns some progress has been made. In particular, multiple psychological studies found flaws in voltional agency, i.e. they point out that we are not always aware of what we do (e.g. we confabulate or make up reasons when we are confused) and why we do it (e.g. being primed). Neuroscientists have also began to make progress by studying free will as “voluntary action”. Several studies found that the timing of a voluntary action are preceded by increases in neural activity even before awareness in humans and before movement in mice. Many of these terms can be objectively defined, tracked and manipulated.

Behaviour prediction as an objective target of free will research


While their interpretation is still being debated, these preliminary findings suggest that otherwise voluntary actions could be decided on long time scales by preconscious processes and predicted prior to our knowledge. Indeed there are some studies that suggest above chance prediction of a human choice up to 10sec prior to awareness of the choice. In parallel, there are non-human animal studies that show that swim direction (zebrafish) or lever-pull time (mice) can also be decoded (or predicted) up to several seconds prior to movement.  Arguably, this suggests that similar or related dynamical systems underlie otherwise voluntary decisions in humans and non-human animals - and that characterizing the anatomy and dynamics of neural processes underling prediction of future behaviours is an objective goal that could be pursued in non-human animals.

As a side note, it should be noted that behaviour prediction is not a new field.  In fact, private companies using artificial intelligence (AI) and social behaviour data have been increasingly able to predict future behaviours of online users (e.g. Facebook algorithms to predict the end of a romantic relationship). Such companies are aggressively pursuing "intent" prediction technologis and patents (Facebook 2018 "life changes" application). As brain recordings from humans become more common it will be exciting and also concerning to see prediction science will continue to grow.