Free Will

Can science explain everything?

 

If by everything, one means the natural physical material world, then perhaps eventually yes, although mysteries might remain. But what about metaphysical entities? Positing that everything only consists of the physical is a philosophical, not a scientific, argument. But science is greedy and constantly attempts to colonize the metaphysical realm. One recent area in which this might be happening is the concept of free will.

 

In Bjorn Brembs’ article (Proc. R. Soc. B, 2011, 278, 930-939), the author begins with several provocative questions.

 

What could possibly get a neurobiologist with no formal training in philosophy beyond a few introductory lectures, to publicly voice his opinion on free will? Even worse, why use empirical, neurobiological evidence mainly from invertebrates to make the case? Surely, the lowly worm, snail or fly cannot be close to something as philosophical today as free will?

 

Brembs weaves an interesting story about adaptive behavior using a range of invertebrates as examples. How do you stay alive and not be eaten by predators? How do you find food when there’s a famine in your vicinity? Turns out that you want to incorporate some degree of randomness into your actions, making it more difficult to predict what you’ll do, and that the actions (or reactions) can be honed by responding to changes in your environment – assuming you’re still alive. There’s an interesting interplay between learning from external stimuli and learning from internal ‘self’ mechanisms. There’s even a part of an insect brain (“mushroom bodies”) that helps control the balance between these two sources.

 

In animals and humans, the situation is both murkier and more complex. We can and do self-initiate actions, our behavior is also dependent on our experiences up to the present moment and on the present, possibly novel, external prompt that might motivate an action. This sense of agency, sidestepping the complicated question of defining consciousness, is what we humans might call free will. Brembs argues against strict determinism or dualism, and instead suggests that we consider free will not (strictly) as a metaphysical identity, but rather as a “quantitative, biological trait, a natural product of physical laws and biological evolution.”

 

Picking up on these ideas, the psychologist Thomas Hills argues for a concept he calls “neurocognitive free will” (Proc. R. Soc. B, 2019, 286, 20190510). To set the stage, Hills defines what he means by conscious control.

 

Conscious control processes are effortful, they focus attention in the face of interference, they experience information in a serial format (one thing at a time), they can generate solutions that are not hard-wired, and they operate over a constrained cognitive workspace – working memory – to which ‘we’ have access and can later report on as a component of conscious awareness. When additional tasks are added to consciously effortful tasks performance suffers. Effortful processes sit in contrast to automatic processes, which are fast and parallel, and do not require conscious awareness. Effortful tasks can be made automatic through repetition (like reading and driving) …

 

Hills assumes that alternative possibilities must be present and able to be acted on for an organism to be ‘free’. Like Brembs, Hills identifies two broad situations where an organism needs to generate such alternatives: exploration and outwitting adversaries. Why and where does this behavioral variability arise? Hill writes:

 

There is a finite precision on cognitive abilities, which is a result of a trade-off between computational accuracy and the metabolic cost of information processing. This can lead to sensory noise, … channel noise, … synaptic noise … Neural systems are commonly characterized as having a sensitive dependence on initial conditions of arbitrarily small size… What matters more for free will is where the decision to modulate variability comes from. If conscious control in any way influences unpredictability, then consciousness is in the loop that governs future behaviour.

 

Some animal experiments are cited, where neural activity involving past experiences is observed even when the external stimuli are no longer present. Apparently this ‘replay’ also happens in dreams. Hills argues that when encountering a ‘choice’, this replay kicks into action by sort-of running a quick (simplified) simulation that takes into account past experience (both good and bad) and exploring different routes. Some of this may be automated or partially automated (I’m assuming), but conscious control is also present and actively involved. In a sense, one predicts what happens to future self in these scenarios. The process may not use all the information streaming in. In fact, conscious control inhibits acting immediately while all this deliberation is taking place. As to the feeling that we have some control in the act of choosing, Hills argues that our ignorance of the future represents the other side of the same coin.

 

… it is exactly the finding out – the initiation of the search and the choice among alternatives – that is the basis of the self’s emergent will and its genuine freedom. The bringing of forth of a self-identity is the evaluation of alternatives through self-simulation. If a historical self emerges through conscious deliberation, and that deliberation involves simulation of alternative futures over which the self chooses, then a historical identity and the capacity for free choice arise in tandem.

 

Could machines have free will? Or at least the ability to “creatively” choose among multiple alternatives? In tandem with Brembs and Hills, the physicist Hans Briegel has an interesting theory which he calls “projective simulation” (Sci. Rep. 2012, 2, 522). First, he tackles the question of why we are reluctant to say machines have free will even though we might ascribe to them some form of intelligence (“the capability of the agent to perceive and act on its environment in a way that maximizes its chances of success”). It’s because the underlying stratum is an algorithm – which is therefore predictable regardless of whether it is deterministic or probabilistic.

 

Briegel has three pillars for his projective simulation. The first is memory – you have to be able to store knowledge of past action. But if memory is all-controlling, there’s no room for variation and adaptation. That’s where randomness comes into play, when it introduces variation at the very point when an organism interacts with its environment. It’s crucial that this randomness be tied to functional ability. Finally, the simulation (with many similarities to what Hill describes) does a random walk through “clips” of episodic memory – a stripped-down version of a detailed simulation. These clips have linkages of different strengths which modulate the probability that the random walker traverses them. But new clips can be created that are not memories but inventions and fabrications, maybe through a mash-up. We can imagine unicorns even if we’ve never seen one. According to Briegel:

 

The fundamental problem is… how freedom can emerge from lawful processes. Both the freedom of self-generated action and the freedom of conscious choice require, at a certain level, some notion of room to manoeuvre, which is consistent with physical law… Room and ultimately freedom arises in two ways, first by the existence of a simulation platform, which enables the agent to detach itself from an immediate (stimulus-reflex type) embedding into its environment and, second, by the constitutive processes of the simulation, which generate a space of possibilities for responding to environmental stimuli. The mechanisms that allow the agent to explore this space of possibilities are based on (irreducible) random processes.

 

All this makes me thinks of games – there are underlying rules, yet the outcome cannot necessarily be pre-determined until the game is actually played. While there are “no-luck” games such as Tic-Tac-Toe where the range of possibilities can be enumerated easily, with more complex and interesting strategy games, the possibilities cannot be computed especially once you throw in dice rolls and/or drawing from a card deck. When I’m playing a game, I try to anticipate what the other players might do. I also have strategies in mind based on previous games I’ve played – those that worked and those that didn’t work. I also have to account for how the current situation on the board may differ from the previous games I’ve played. I’m not sure how exactly I compute all these possibilities, but eventually it’s my turn and I make my move. I don’t suffer from “analysis-paralysis” when playing a game, but maybe it’s because I’m not sufficiently patient, or alternatively maybe because I’m generally decisive.

 

If I had a computer app associated with a game I’m playing, would I use it as an aid? I don’t know since I’ve never tried it personally. I don’t play games on the computer since I already stare at the screen for many hours when I’m at work. But I could imagine that if you’re playing a computer game, you could have an app that does some predictive simulation based on what has unfolded so far in the present game, while also feeding in information from previous games – basically an algorithm that chunks through data. But while that might make you better informed, the result of the game is still open and you’ll have to play to the finish. Perhaps that’s akin to the “freedom” alluded to by the authors I mentioned in today’s post. I certainly feel that I have free will when making at least some of my choices. But I don’t doubt that unconscious factors come into play in every choice that I make.