Could God See Freewill?
(Annotated by Galen Strawson - GS)
The problem of free will becomes especially sharp when placed alongside modern physics and contemporary philosophy of language. If the laws of nature together with the complete state of the universe at some past time fix everything that ever happens, then in what sense could anyone have done otherwise? Yet at the same time, developments in epistemology and metaphysics suggest that our experience of openness may reflect deep features of our epistemic situation rather than metaphysical indeterminacy.
To understand this landscape, I propose to bring together four figures whose work illuminates different aspects of the issue: Tim Williamson on epistemic vagueness, Roy Sorensen on blindspot ignorance, Barry Loewer on the Mentaculus and objective chance, and Kit Fine on hyperintensional modality. We can then place their positions within a broader framework inspired by Rudolf Carnap and later refined by David Chalmers, before finally asking what an omnipotent God would know, and whether such a being would “see” free will. [fashionable references! … my inclination is to bring in historical figures, who have almost always made the relevant points, usually better than our contemporaries -GS]
The classical threat to free will is captured by what is often called the Consequence Argument, most prominently associated with Peter van Inwagen. Suppose determinism is true. That means that the laws of nature and the complete state of the world at some time in the distant past together entail every later event. Your present decision, on this picture, is the consequence of facts long before you were born and of laws you did not choose. You cannot change the past, and you cannot change the laws. So it seems you cannot change their consequences. If free will requires that you could have done otherwise in a strong sense, determinism appears to undermine it. [point made 1000s of times before Peter van Inwagen. Somewhat absurd that this point associated particularly with his name—I’m inclined to be generally grumpy about philosophy’s forgetting of its past—bit like the fish played by Ellen de Generes in Finding Nemo -GS]
Much of the debate therefore turns on what “could have done otherwise” means [discussed to death in the past …. Including G. E. Moore J. L. Austin but also 100 others -GS]. To analyse such claims, philosophers turn to counterfactuals, statements of the form “If I had done X, Y would have happened.” One influential approach to counterfactuals was developed by David Lewis, who suggested that such statements are true when, in the most similar possible worlds where X occurs, Y also occurs. Barry Loewer argues that this similarity-based framework fails once we take the physics of statistical mechanics seriously. He proposes instead the Mentaculus framework.
The Mentaculus consists of three elements: the fundamental dynamical laws of physics, a special low-entropy initial condition known as the Past Hypothesis (David Albert first developed this I think), and a probability distribution over microstates compatible with that initial condition. The Past Hypothesis explains the arrow of time: why entropy increases, why eggs fry but do not unfry, why memory points toward the past. Given the laws and this special beginning, statistical mechanics yields an objective probability distribution over possible micro-histories of the universe. Even if the underlying laws are deterministic, these probabilities are not merely subjective measures of ignorance. They arise from the structure of the physical theory and the coarse-grained relation between microstates and macrostates.
Loewer uses this framework to reinterpret counterfactuals. Instead of asking which possible world is most similar to the actual world, he asks what the objective probability of Y is given X and the relevant macroscopic state, according to the Mentaculus. At the microscopic level, determinism may fix a single history. But at the macroscopic level, many microstates correspond to the same macrostate, and these can evolve differently. From the perspective of macroscopic agents, the future is structured by objective probabilities. On this view, if you had chosen differently, the microscopic history of the universe would have differed in ways that leave the macroscopic past intact. The Consequence Argument’s appeal to an entirely fixed past is weakened once we distinguish between micro and macro levels. [I wonder …GS]
Against this backdrop, Tim Williamson’s epistemicism introduces a different way of thinking about apparent openness [crazy view! -GS]. Williamson argues that vagueness does not reflect indeterminacy in reality but ignorance of sharp boundaries. There is a precise fact about when a heap ceases to be a heap, but we cannot know it. Reality is sharp; our knowledge is limited. Applied to free will, this suggests that even if the future feels open, there may be a precise microphysical fact that fixes what will happen. The branching we experience may be epistemic rather than metaphysical. The world itself may be entirely determinate at its base, and our sense of alternatives may reflect ignorance of that base.
Roy Sorensen’s work on blindspots deepens this epistemic dimension. Sorensen identifies truths that cannot be known, not merely because of limited evidence, but because knowing them would undermine their truth. For example, you can’t know the conjunction “p and I don’t know p”, where p is the specific truth in question. (I corrected this having made a faux pas in the original draft - thanks for Tim Williamson pointing it out!!) When applied to free will, Sorensen’s insight suggests that even in a deterministic universe, agents embedded within the system cannot access the complete microphysical description. [old point … see e.g. D. M. McKay 1960 (‘On the Logical Indeterminacy of a Free Choice.’ Mind 69: 31-40), Popper, Pears, ‘Predicting and Deciding’, -GS] There may be truths about their future actions that they cannot know in principle. The openness of the future, from their perspective, may therefore reflect structural features of being situated within the world.
Kit Fine’s hyperintensional approach complicates matters further. Fine argues that modal and counterfactual claims are more fine-grained than standard possible-worlds semantics allows. Two statements can be true in exactly the same possible worlds yet differ in their content because they differ in grounding or explanatory structure. The phrase “could have done otherwise” may express distinct claims. It might mean that there exists a physically possible world with the same total past and laws in which one acts differently. Or it might mean that, given one’s capacities, reasons, and macroscopic conditions, an alternative was available in the sense relevant to control and responsibility. These notions are not equivalent. [ancient point, made 100s of times -GS] If one side in the debate insists on the first reading and the other on the second, the disagreement may be partly verbal or framework-dependent.
At this point, the work of Rudolf Carnap and David Chalmers provides a broader interpretive frame. In The Logical Structure of the World, Carnap sought to show how the world could be constructed from a minimal base via structural relations. He was less concerned with discovering the ultimate furniture of reality than with showing how different levels of description can be systematically organised. David Chalmers, especially in his work on scrutability and metametaphysics, extends this approach. He argues that many metaphysical disputes concern which conceptual framework we adopt. Once a base theory is fixed, higher-level truths can in principle be derived or “scrutable” from it plus appropriate conceptual analysis. Imagine, then, that we could in theory map out the entire universe. We have the complete microphysical description, the laws, the Past Hypothesis, every particle position and momentum. From a Carnap–Chalmers perspective, this constitutes a base theory. The question is whether free will survives as a higher-level construction from that base. [pretty sketchy it seems to me - GS]
If the base theory entails that agents deliberate, are responsive to reasons, and satisfy certain counterfactual conditions [say which! - GS] , then free will may be constructed as a higher-level structural fact. It would not compete with the microdescription but be grounded in it. The fact that the base is deterministic does not automatically eliminate higher-level truths. Thermodynamics survives microphysics; psychology survives neuroscience. On this model, free will survives determinism if it is defined in terms of macro-level control and objective probabilistic structure. [I think you mean moral-responsibility-entailing free will? -GS]
Williamson’s sharpness at the base level does not eliminate this possibility. Sorensen’s blindspots reinforce it by highlighting that agents cannot occupy the God’s-eye perspective of the full base description. Fine’s hyperintensional distinctions caution us against assuming that only one notion of possibility matters. The Carnap–Chalmers framework suggests that the debate may concern which level of construction is metaphysically relevant for freedom.
This brings us to a final, speculative question: what would an omnipotent God or Laplacean demon know, and would such a being “see” free will? Suppose God or the demon has complete knowledge of the microphysical base, the laws, the Past Hypothesis, and every true proposition. On a deterministic universe, God would know the entire future with certainty. From the divine perspective, there would be no epistemic openness. Every action would be known as a consequence of the total state and laws. Does that mean God would not “see” free will?
Not necessarily. God’s knowledge would include not only the base facts but also all higher-level truths constructed from them. If free will is a higher-level structural feature grounded in macro-level probabilistic relations and rational responsiveness, then God would know that as well. God would know that agents are responsive to reasons, that their actions track their deliberations, that certain counterfactuals are true within the relevant framework. From the divine perspective, the openness of the future would not be epistemic, but the structure of agency would still be real [the structure of intentional agency is certainly compatible with determinism - GS].
God would not see metaphysical indeterminacy at the base, but might see a layered reality in which higher-level structures, including agency and responsibility, are fully grounded. [responsibility — but what kind? - GS] If, however, free will requires fundamental metaphysical indeterminism at the base level, then a deterministic micro-map would eliminate it, even for God. In that case, God would see only a single fixed history unfolding. [it’s an ancient point that indeterminism can’t help with radical freedom responsibility in any way - GS]
So maybe free will does not require metaphysical openness at the base level. [perhaps see R. E. Hobart 1934?] What matters for responsibility and agency is macro-level control, rational responsiveness, and the truth of certain counterfactuals within the Mentaculus framework. A fully mapped universe would fix the base, but it would not erase the layered structures constructed upon it (including our best physics). From the divine perspective, there would be no ignorance, no blindspots, no epistemic openness. Yet the structural features that constitute agency [again, I think you need to define agency, in particular free agency … dogs are agents- GS] would still be present in the constructed hierarchy of truths. [agency as ordinarily understood is fully compatible with determinism without reference to any of these suggested complications - GS ]
On this view, free will is neither a mysterious metaphysical gap nor a mere illusion born of ignorance. It is a higher-level structural feature of complex agents in a probabilistically organised macroscopic world. [but why does this help with responsibility?] An omnipotent God would not see the future as open, but would still see agents as free in the only sense that matters for moral life: as systems whose actions flow from their reasons, capacities, and character [that is certainly what we are - GS] within the layered architecture of reality.[again don’t see how any of the fancy Barry Loewer stuff makes any difference! - GS]
Some Philosophers Comment Further
Of God, Ham Sandwiches and Freedom of the Will
It is noon and I am hungry. I go to the kitchen to make a sandwich. I have other options: there is soup, there are cookies. I opt to make a ham sandwich. My physical state caused my hunger; I did not choose to be hungry. Likewise, I am confident that the molecules in my brain caused me to make a ham sandwich. Hunger happened to me. I didn’t do it.
Making the ham sandwich is something I do, not something that happens to me. That difference is as vivid to me and as indubitable as is feeling hungry or feeling pain. I feel myself as the very source of the choices and actions. I am certain I could have chosen soup or cookies or a tuna sandwich. How then could I have done differently if, as I also believe, my brain determined my choices and my brain is just a physical system working according to physical laws, and earlier physical states determined, as much as anything can, my brain states at the moments when I chose? I could not have. So I believe a contradiction.
The early Christian fathers and many of their later followers worried about how God’s omniscience could be reconciled with freedom of individual will. Their solutions turned on fiddling with the meaning of “free” or “will.” To solve my problem, contemporary philosophers do the same with “choose” and “could have” and “free to.” Some tell me one of my beliefs is true of little things and the other of big things, like Man’s level and God’s level. I am grateful for their efforts on my behalf, but they are unavailing. Until the scientists can show me how my indubitable feeling arises from the molecules, I will live discontentedly with the contradiction and read again Of Human Bondage—Maugham, not Spinoza.
(Clark Glymour Alumni University Professor Emeritus, Department of Philosophy, CMU)
3 Points
At least at three points in “Could God See Free Will” Richard Marshall adverts to the role of reasons in agents’ deliberations as the relevant source of free will—the feature of human behavior that constitutes the exercise of our free will:
[1]If free will is a higher-level structural feature grounded in macro-level probabilistic relations and rational responsiveness, then God would know that as well. God would know that agents are responsive to reasons…
If deterministic basement level processes ground higher level “structural facts” and deliberation constituted such a structural fact, then determinism at the base need not generate determinism at the higher levels:
“The fact that the base is deterministic does not automatically eliminate higher-level truths. Thermodynamics survives microphysics; psychology survives neuroscience.”
One might wonder whether 2d law objective chance and human intentionally are emergent from deterministic microphysics in exactly the same ways. But leave this issue aside.
Free will, Marshall suggests, might be a higher-level structural feature of complex agents in a probabilistically driven macroscopic world. Whence an affirmative answer to his question:
“An omnipotent God would not see the future as open, but would still see agents as free in the only sense that matters for moral life: as systems whose actions flow from their reasons, capacities, and character within the layered architecture of reality.” [Emphasis added]
But, think about the italicized qualification above. One needs to ask what’s so special about reasons and rational deliberation as a foundation for attributions of free will? This was the question Galen Strawson posed some time ago in “Luck Swallows Everything.” It might be a matter of interest in the debate between dualism and physicalism that reasons are as emergent from a deterministic micro base as objective chance, but by itself this won’t do much for conferring upon them what is needed for a free will worth wanting, in Dan Dennett’s haunting phrase, what matters for moral life.
(Alex Rosenberg The R. Taylor Cole Professor of Philosophy, Duke University)
Reality and Appearance
In 2021, I published Appearance in Reality (Oxford) the focus of which was on the chasm between the world as it appears to us (Wilfrid Sellars’ manifest image) and the world according to physics (Sellars’ scientific image). At one time physicists held that the cosmos was a spacetime ‘block’. Today physicists struggle to reconcile quantum mechanics and general relativity, and could be headed toward the elimination of space and time altogether. Whatever the upshot, the contrast between the manifest and scientific images could not be more stark.
Suppose that the world as described by physics is an unchanging four–dimensional block, or maybe one that elides space and time altogether. In either case, nothing changes, nothing moves. Motion would resemble the ‘motion’ of a cricket ball across a television screen. Would this mean that motion, change, and the like are illusions? That would seem to undermine the evidence we have for the theories: instruments deployed by the sciences have moving parts and keep track of changes over time. Another possibility is that physics and its sister sciences are meant only as guides for navigating the appearances, and not meant to be taken as accounts of reality.
A third, currently popular, option is that reality is hierarchical: the appearances are ‘higher-level’ phenomena dependent on but distinct from phenomena at lower levels. I have spent much of my time in recent years arguing that none of these options is viable. Rather, physics, the scientific image, provides our best accounting at any given time as to the character of truthmakers for truths at home in the manifest image: temporal truths could have atemporal truthmakers, truths about responsibility and free action could have deterministic truthmakers. The manifest and scientific images afford different, complementary ways of talking about – ‘carving up’ – the world, different taxonomies, the mistake is to assume that these are in competition. Physics affords an account of the nature of the truthmakers for truths concerning the appearances.
(John Heil FAHA, Professor of Philosophy Emeritus Durham University and Washington University)
The Weather Vane
I am a compatibilist about free will. Determinism, were it true, would not rob us of free will, and indeterminism would not secure it. Until 1992, I was a compatibilist without a clue as to what a good theory of freewill would be. That changed when my colleagues and friends, Dennis Stampe and Martha Gibson, published “Of One’s Own Free Will” (Philosophy and Phenomenological Research, Vol. 52, 1992, pp. 529–556). Their paper makes use of a brilliant analogy. True, an analogy isn’t a theory, but I think their analogy points in the right direction for finding the theory we need.
Consider a weather vane. Its function is to indicate the direction in which the wind is blowing. A weather vane that is stuck can’t do this. What the weather vane needs, if it is to perform its function, is to be free. This doesn’t mean that its pointings are uncaused. On the contrary, it means that they are. A free weather vane is caused to point in the direction the wind is blowing; a weather vane that can’t swivel is unfree, though it may fortuitously be pointing in the right direction. Weather vanes have the function of providing information about the wind’s direction because they were built by human beings to do so.
The will also has a function. Its function is to take account of the agent’s beliefs and desires and formulate intentions to act. The will takes desires and beliefs as inputs and outputs intentions. For the will to be free, the desire-generating device and the belief-generating device must be functioning properly – they need to be reliable. The desire-generating device has the function of telling an agent what would be good for her to do; the belief-generating device has the function of telling an agent what is probably true. Free will diminishes or disappears when individuals are psychologically “stuck.” This is what happens with addictions, and other types of psychological impairments.
(Elliott Sober William H. Bonsall Visiting Professor in the Humanities, Department of Philosophy, Stanford University; Hans Reichenbach Professor and William F. Vilas Research Professor, emeritus, Philosophy Department, University of Wisconsin – Madison. https://sober.philosophy.wisc.edu/)
Determinism's Not The Problem
Peter van Inwagen thinks that free will is threatened by determinism. As he sees it, if past facts plus the laws of nature logically entail what we do, then we can’t be in control of our actions. However determinism can’t really be the problem. After all, on standard versions of physical determinism, future facts logically entail what we do, just as much as past facts. Given the physical laws of nature, my house being newly painted in April logically entails that I had it painted in March. Yet, nobody, I take it, feels that this threatens my freedom. On the contrary, it’s precisely because I was free to have my house painted in March that it ends up sparkling bright in April.
In truth, it’s not determinism that we should be worried about, but causation. Our freedom is certainly threatened by causal constraints, even if it is left untouched by physical determinism. But that is a different matter. As Jenann Ismael explains in her fine book How Physics Makes Us Free, causation is a macroscopic phenomenon, peculiar to large-scale objects and events, and as such transcends the deterministic patterns of basic physics. When the world reaches heat death, physical determinism will survive, but there will be no causation. So even if all things, including our actions, are physically determined, it does follow that they are always causally constrained.
Having said that, I don’t want to deny that our actions often are causally constrained, as well as determined, and moreover in ways that do undermine our freedom. Someone who is locked up in prison isn’t free to go to the beach. It is nice question how far the less coercive causal influences to which we are all subject similarly take away our ability to choose freely. If you ask me, sometimes we’re rendered unfree, sometimes not. But in any case that’s the question to focus on. We should put the bogey of determinism to one side, and instead think hard about the different ways in which our causal histories can limit our choices.
(David Papineau, Professor of PhilosophyKing’s College London)
I have doubts about the soundness of Sorensen’s argument that there is a limit to what can be known because there are truths that cannot be known since knowing them would undermine their truth. Well, I at least have doubts about the particular example given of such an unknowable truth, i.e. the conjunction “p and I don’t know p”. To begin with, I am struggling to spell out for myself what the relevant concepts of truth and knowledge are supposed to be from which this conclusion follows. I can kind of understand it if we assume that there just are primitive ‘truths’ which we than know if we know the truths (rather than knowing the facts that make them true). But I am not sure I could pin such an understanding of truth and knowledge on any particular philosopher. Under the assumption of a more widely accepted alternative today would be that we are dealing with two propositions “p is the case” and “I do not know that P is the case”, which would be true if and only if there obtain certain facts that make them true. These would arguably be that p is the case and that whoever ‘I’ refers to—say, ‘Peter’—does not know that p is the case. Now, why would it be impossible for Peter to know “p and I don’t know p”? I would say it is because as soon as Peter knows p then “I do not know p” is no longer a truth, and that does not seem like a limitation in Peter’s knowledge in those circumstances that he couldn’t know that he doesn’t know p. Indeed, it seems the preconditions for Peter to know “p and I don’t know p” is for the truth-value of the proposition to remain unchanged also when the truth-maker for that proposition no longer obtains (Peter being ignorant of p). In conclusion, it seems this particular example does not show that there are truths that cannot be known (in any sense that would indicate a limit on knowledge) but rather that there are truths that cannot remain truths if we know them, notably in the cases where the precondition for their truth is our ignorance about them. That those truths remain no longer true when we no longer are ignorant, seems like a good conclusion and not one that indicates a limit to our knowledge.
(R.D. Ingthorsson, Dept. of Philosophy, History, and Art Studies
University of Helsinki)