Department of Astronomy Harvard University

Information Philosopher Social Trends Institute

The Two-Stage Solution to the Problem of Free Will

Reconciling Freedom With a Limited Indeterminism and a Limited, but Adequate, Determinism

Department of AstronomyHarvard University

Bob DoyleInformation Philosopher


1950'sWhen I went to Harvard to study astrophysics in 1958, the Big Bang theory said the universe began in a state of equilibrium 10 or 20 billion years ago.

Given the second law of thermodynamics, I asked, why isn't it still in equilibrium?.

Why do we have a cosmos, rather than chaos? Why something, rather than nothing?


1960'sMy Ph.D thesis was on the interaction of two (entangled)hydrogen atoms with radiation (photons). I found the universe to be discrete, not continuous, quantum rather than classical, digital, rather than analog.

I read Arthur Stanley Eddington, who said that quantum mechanics shows that pre-determinism is not true. Like Eddington, I connected quantum physics to free will.

Quantum randomness generates new possibilities for the universe, I thought, like the evolution of biological species.


1970's – DennettIn 1978, Dan Dennett described a two-stage model for free will – a production stage with a random element and a subsequent choice stage.

But he thought only deterministic pseudo-random numbers were needed to generate the alternative possibilities.

At the time, I thought Libertarians would soon accept his two-stage model, but of course with indeterministic quantum randomness to break the deterministic causal chain back to the origin of the universe.


1980's – KaneToday's leading Libertarian, Bob Kane, has always accepted quantum randomness. But he locates indeterminism in the "plural rational control" decisions needed for self-forming actions (SFAs) and "ultimate responsibility" (UR).

He said that Dennett's model was "a significant piece in the overall puzzle," suitable for practical reasoning if not moral, but given the randomly generated possibilities, an agent would be determined to choose the best available option.

Perhaps, but the agent would not be pre-determined, even from moments just before deliberations began.


1990's – MeleAl Mele argued for a "modest libertarianism" that is essentially the Dennett decision-making model. But he remains agnostic on the role of quantum mechanics – for him determinism might still be true. He says,

"It might be worth exploring the possibility of combining a compatibilist conception of the later parts of a process…with an incompatibilist conception of the earlier parts…Compatibilists may, in principle be willing to accept an account of causation that accommodates both deterministic and probabilistic instances." (Autonomous Agents, pp.212-13)

I agree. I hope Mele is right about compatibilists agreeing.


2000's – HeisenbergIn the May 14, 2009 issue of Nature, Martin Heisenberg argued for behavioral freedom in lower animals, based on a two-stage process of randomness followed by lawful behavior. He felt it could be the basis for free will in humans.

I wrote Nature to say that a two-stage model had been first suggested by William James in 1884. Since James, it was also discussed by Henri Poincaré, Arthur Holly Compton, Karl Popper, Henry Margenau, John Martin Fischer, Stephen Kosslyn, and, of course, Dennett, Kane, Mele, and myself.

It is now the most practical and plausible model for free will.


My Two-Stage Model for Free Will

Stage 1) Alternative possibilities generated by chance, in part by quantum events that break the causal chain of pre-determinism.

Stage 2) An adequately determined evaluation of those alternatives, resulting in a willed decision.

First chance, then choice. First spontaneous variation, then selection (Darwin analogy).First possibilities, then actuality.Thoughts come to us freely. Actions go from us willfully. First “free,” then “will.”

Information Philosopher Social Trends Institute04/20/23 9

Dennett's Objection

"Isn't it the case that the new improved proposed model for human deliberation can do as well with a random-but-deterministic generation process as with a causally undetermined process? " (Brainstorms, p.298)

My reply: Quantum randomness is needed tobreak the causal chain of determinism.It expands the range of creative alternative possibilities.


Kane's Objection "[The agent] does not have complete control over what chance

images and other thoughts enter his mind or influence his deliberation. They simply come as they please...What happens from then on, how he reacts, is determined by desires and beliefs he already has. " (A Contemporary Introduction to Free Will, p.64-5)

My reply: Determination by reasons is not pre-determined, as R.E.Hobart pointed out in 1934.

Libertarianism need not require that the will itself be free and undetermined by reasons, in the sense of quantum randomness.


Mele's ObjectionMele argues that true randomness makes options a matter of luck and

luck might compromise the control needed for moral responsibility.

"When luck (good or bad) is problematic, that is because it seems significantly to impede agents' control over themselves." (Free Will and Luck, p.7)

My reply: Luck is the simple consequence of real chance in the universe. As long as our evaluation and selection of options are adequately determined by our reasons, motives, and desires, we are responsible, however the options were generated.


Martin Heisenberg's Behavioral Freedom

Selection methods get us from Heisenberg’s behavioral freedom in lower animals to my free will in humans.

Whether in biological evolution or in generating possibilities for our two-stage model, the source for spontaneous variations, is always the same – indeterminism, objective chance, quantum randomness.

It is the method of selection that makes the difference.

In biological evolution, nature selects.In behavioral freedom. the organism itself “purposely” and "intentionally" selects.


Four Levels of Selection

Instinctive selection - selection criteria are transmitted genetically, shaped only by ancestors' experiences. Learned selection - for animals whose past experiences guide current choices. Selection criteria are acquired through experience, including instruction by parents and peers. Predictive selection - using imagination and foresight to estimate the future consequences of choices. Reflective (normative) selection – conscious deliberation about community values influences the choice of behaviors.


Comparing My Two-Stage Model

My model is Dennett's model, plus genuine quantum randomness in the generation of alternative possibilities.

Kane has long held that quantum randomness is needed. But my model puts chance in the early stage, not in the decision.

I agree with Mele's views, except that determinism is false. Today I will provide a second reason why it is false.

None of these thinkers locates the time and place of the sources of quantum randomness as I do.


What's Better About My Two-Stage Model?

Previous thinkers could not amplify a single quantum event in the brain and synchronize it to make a decision free (uncaused), yet provide agent control and responsibility.

My model does not rely on a single quantum event for each willed decision. That would make the decision random.

The source of randomness in my model is the ever-present quantal and thermal noise that affects the creation, storage, maintenance, and retrieval of information in any information-processing system.


Noise In All Information Processing

The source of a break in the causal deterministic chain need not happen in the mind itself. It could be external.

And it need not happen during deliberation. It just needs to be considered during deliberation.

It could result from a mistaken perception, or an error during consolidation of memory, or noise in memory retrieval.

Creating information always involves quantum indeterminacy.


The Origin of Information

To return to the original question, given the second law of thermodynamics, why isn't the universe still in equilibrium?

How can we be having this conversation and exchanging new information? Because we, and the universe, are creative.

We humans create new information, using the same steps as what I call the cosmic information creation process.

It involves the interaction of quantum physics (with its indeterminacy) and thermodynamics (with its irreversibility).


The Cosmic Creation ProcessInformation creation requires two things – a microscopic quantum event that lowers the entropy locally, and energy and entropy transfer away from the new information structure, needed to increase entropy globally.

Quantum processes produce all bound states of matter – fundamental particles, atoms, molecules, DNA, etc.

But binding energy, and thermodynamic entropy, must be carried away from any new information structure for it to be stable. Otherwise, it will be unstable and disappear.


The Cosmic Creation Process

Let's form a hydrogen atom from a proton and an electron.

p+ + e- -> H + ν

The binding energy is released from the hydrogen atom as a photon (ν), which is radiated away.

If that energy and the associated thermodynamic entropy is not carried away from the hydrogen atom, it will separate back into a proton and electron (absorbing a photon).

We can reduce the entropy locally, if we increase it globally.


Time and Information Creation

Whether it is a new particle like a hydrogen atom,or a bit set or reset in a digital computer,whether it is the next nucleotide attaching to a DNA strand,or a new idea for an alternative possibility in your mind,two irreversible things must happen…

An undetermined and irreversible quantum event.

The irreversible transfer of entropy away from the new information structure.

Irreversibility in information creation is the arrow of time.


Entropy and InformationLeo Szilard in 1939 showed that Maxwell's Demon must gather information to decrease the entropy.

Gunter Ludwig in 1953 showed that in the measurement process, for every bit of information acquired, at least one bit of entropy is carried away by the measuring apparatus.

Rolf Landauer (1961) showed that the change of one bit of computer data requires the computer system to absorb one (usually many more) bits of positive entropy.

But how can information (good negative entropy) possibly increase at the same time that (bad) entropy increases?


Information in Classical Physics

Laplace's Demon (1814)

A Laplace Demon has all the information - forces, positions, velocities - for all the particles in the universe.

All times, past and future, are present to the Laplace Demon,as to the eyes of God. In a deterministic universe, information is constant.

Mathematical physicists, like Laplace, believe that the conservation of informationis as much a conservation law as those of matter and energy.

There is no chance. The randomness we see is simply epistemic, a consequence of human ignorance about physical details that a demon and God can know.

information a constant of nature

info

rmat

ion

time(Pierre Simon Laplace, A Philosophical Essay on Probabilities, 1814)


Entropy As Lost Information

(William Thomson, "On a Universal Tendency in Nature to the Dissipation of Mechanical Energy")

Following the discovery of the laws of thermodynamics, William Thomson (Lord Kelvin) claimed that the universe would “run down,” all the energy ultimately dissipated into thermal motions,which Herman Helmholtz called a “heat death.”

Mathematicians would say the information lost to entropy is stillavailable microscopically, and recoverable if time was reversed.

Lord Kelvin’s Heat Death (1852)

entr

opy/

info

rmat

ion

entropy

information


Making Room for Information

As the universe expands, the number of phase space cells and the maximum possible entropy expand much faster than the matter and energy can re-equilibrate (reach thermal equilibrium), leaving room for negative entropy, and for stable information structures to form and grow.

actual entropy

maximum possible entropyinformation

negative entropy

time

entr

opy

/ inf

orm

atio

n

David Layzer, The Arrow of Time, Scientific American, 1975

GOOD?

BAD?


Now Two Reasons To Reject Determinism

Old reason: information creation requires quantum events, which are inherently indeterministic. The future is only probabilistic, though it may be "adequately determined."

My new reason: there is not enough information in the past (none at all in the early universe) to determine the present.

The "fixed past" and the "laws of nature" pre-determine nothing, despite recent philosophical arguments.

Similarly, information at the present time does not determine the future. The future is open. We must create it.


Resolving Two Paradoxes

Ludwig Boltzmann's ideas on the increase of entropy with timewere attacked using classical mechanical ideas which now fail.

The first was Loschmidt's paradox. If the velocities of particles could be turned exactly around, it would be as if time had reversed and the entropy would go back to its initial value.

The second was Zermelo's paradox. Given enough time, any given distribution of particles among all the phase space cells would beexactly repeated. Again, entropy should decrease at that time.

Quantum mechanics and information creation resolve these paradoxes.


No Macroscopic Recurrence

The ancient idea of a great year or eternal return is impossible in a universe with increasing information content.

Max Planck's student Ernst Zermelo (who later developed axiomatic set theory) argued that given a long enough time, the particles of any gaswould return to a distribution in "phase space" (a 6n dimensional space of possible velocities and positions) that would be indistinguishable from the original low entropy distribution.

This is true in a closed chemical "universe." But in the open astronomicaluniverse, information increases. Later times can thus never be"exactly the same circumstances."

Compare our human memory of past circumstances.


No Microscopic Reversibility

A careful quantum analysis shows that reversibility fails even in the simplest conditions - the case of two particles in collision.

When they collide, they should not be treated as individual particleswith single-particle wave functions, but as a single system with a two-particle wave function, because they are now entangled.

Molecular (not atomic) wave functions should be used.

Quantum transitions between closely spaced rotational and vibrational energy levels in the "quasi-molecule" introduces uncertainty. Transitions could be different for a hypothetical reversed path.


The Problem of Measurement SolvedIn the deterministic evolution of the Schrödinger equation, information is conserved, just as in classical physics.

Unless there is an indeterministic “collapse” of the wave function, unless at least one bit of information is acquired, and unless at least one (and generally many more) bits of entropy are carried away by the measuring apparatus, there can be nothing for a “conscious observer” to observe.

The Heisenberg/Von Neumann “cut,” John Bell’s “shifty split,” happens at the moment that stable information enters the world.

Knowing this can help us solve the puzzle of Schrödinger's Cat.


John Bell’s “Shifty Split”


John Bell’s “Shifty Split” Located

The spots on thephotographic plate

are irreversiblenew informationin the universe.

Without them, thereis nothing for a

“conscious observer”to observe.


How Quantum Underlies Classical

The Correspondence Principle ensures that the laws of quantum physics asymptotically approach the laws of classical physics in the limit of large quantum numbers.

Similarly, the Law of Large Numbers of statistical events ensures that, normally, microscopic quantum events average out for many- particle systems to produce regular, though statistical, laws.

The world thus shows us an "adequate determinism" for large objects that in no way denies the fundamental and irreducible indeterminacy at the quantum level.

Macroscopic systems such as biological organisms and computers can reliably suppress quantum indeterminism, if they want to.


Dennett's Challenge

Dan Dennett challenged me to give reasons why quantum indeterminacy is better than computer pseudo-randomness.

Dennett does not deny quantum indeterminacy. He just doubts quantum randomness is necessary for free will.

First, quantum randomness has been available to evolving species for billions of years before pseudo-randomness. Second, I proposed four cases where quantum chance is critically important and better than pseudo-randomness.


Laplace Demon?

As we saw earlier, we now know that a Laplace Demon cannot exist in the real universe.

There is not enough information in the past to determine the present, let alone completely determine the future.

Creating future information requires quantum events, which are inherently indeterministic. The future is only probable, though it may be "adequately determined."


Intelligent Designer?

Suppose determinism is true, and that the chance driving spontaneous variation of the gene pool is merely epistemic, in actuality the result of Dennett's computer generated pseudo-random number generator.

An intelligent designer with an even bigger computer might reverse engineer and alter the algorithm behind the pseudo-randomness driving evolution.

But cosmic rays, which are inherently indeterministic quantum events, damage the DNA to produce mutations. No intelligent designer could control such evolution.


Frankfurt Controllers

For the last few decades, compatibilists have used Frankfurt Cases to show that free will does not require alternative possibilities.

But, as Bob Kane first showed in 1985, if a choice is undetermined, the Frankfurt controller cannot tell until the choice is made whether the agent will do A or do otherwise.

Because chance (quantum randomness) helps generate the alternate possibilities, information about the choice does not come into the universe until the choice has been made.


Dennett's Eavesdropper

I call this Dennett's Eavesdropper, because in a discussion of quantum cryptography, Dennett agreed there is a strong reason to prefer quantum randomness to pseudo-randomness for encrypting secure messages.

He sees that if a pseudo-random number sequence were used, a clever eavesdropper might discover the algorithm behind it and thus be able to decode the message.

I am a visitor in Dennett's Free Will Seminar at Tufts this Fall.With luck, he will soon accept more indeterminism in the early generation stage of his model.


Why Information Philosopher?Information is neither Matter nor Energy.But it needs Matter for Embodiment.And it needs Energy for Communication.

I think of information as immaterial, a spirit…Information is the mind in the body.Information is the ghost in the machine.Information is the soul in the flesh.

When we die, it’s our information that is lost.


So Is Science Compatible With Freedom?Free Will is Incompatible with Pre-determinism and with Indeterminism in the Choice itself (excepting Bob Kane’s "torn decisions") .

Free Will is Compatible with Limited Indeterminism and with Limited, but Adequate, Determinism (i.e., determination by reasons, values, and desires).

David Hume reconciled freedom with determinism. I hope to be remembered as reconciling free will with indeterminism

Thank you.


• The following slides are for discussions.


What If Decisions Had Just One Stage?

Determinist philosophers say an action could not have been otherwise, given the “laws of nature” and the “fixed past,” the exact circumstances immediately preceding the decision.

This is because, for them, the decision is a single point in time.


We Need Time To “Do Otherwise”

In our two-stage model, the decision is a process with a temporal sequence, first “free,” then “will.”

Our thoughts come to us freely (generated).

Our actions go from us willfully (evaluated).


We Can Even Have “Second Thoughts”

Our decisions are not necessarily determined once we generate the alternative possibilities.

If our evaluation finds the alternative possibilities unacceptable, and if time permits, we can always generate more creative ideas.


Multiple Causes in the Global WorkspaceBernard Baars’ audience in his Theater of Consciousness ≈Dan Dennett’s functional homunculi with their causal chains


Multiple Causes in the Global WorkspaceBob Kane’s Self-Forming Actions add their own new causal chains


Multiple Causes in the Global WorkspaceMy Two-Stage Model adds new alternative possibilities, - after the Circumstance and before the Decision


Giving Compatibilists What They Need

Given a logical choice between determinism and indeterminism,compatibilists understandably choose determinism, so that their decisions are "determined" by evaluations of their reasons, motive, and desires, in short, by their character.

The Two-Stage Model provides all the "determination" of the will compatibilists want and need, but not the "pre-determinism" that threatens agent control and responsibility.

Might compatibilists accept the limited indeterminism that we have in quantum physics and the real world?


Random Quantum Events in the Brain?

Molecular biologists are skeptical about any quantum indeterminacy in the brain-mind. Neurons are macroscopic objects with the order of 1020 atoms. How could one atom affect anything?, they ask.

Apart from the fact that there are trillions of quantum events in the brain every second, we can also note that biological systems have evolved to the quantum limit. An eye can detect a single photon. A nose can smell a single molecule.

I argue that the brain has found an evolutionary advantage in quantum indeterminacy and thermal noise.


Quantum Mechanics and Free Will

Please note that quantum mechanics contributes more than just indeterminacy to the two-stage model of free will.

Critically, it also provides the stable information structures that the will uses to recruit and manage the indeterminacy.

The stability of information in DNA over cosmic time scales is a direct consequence of quantum discreteness.

Without quantum mechanics, electrons would spiral into the nucleus. Matter as we know it would disappear in an instant.Protons and electrons would become neutrons, not atoms.


Creativity and Free Will

Normally random noise is the enemy of information, but it can be the friend of freedom and creativity. Alternative possibilities are the source of human creativity. They make us the authors of our lives.

We normally suppress the creative noise.

We are perhaps most free when we dream, when we are imaginative, when we are creative.

And when we create, we are co-creators of our universe.


Comprehensive Compatibilism?

For compatibilists, the two-stage model is a Comprehensive Compatibilism because it is Compatible with both…

the Adequate Determinism we need for determination by reasons, values, and desires - which compatibilists want,

and the Limited Indeterminism we need to generate alternative possibilities. - which compatibilists might accept.


David Hume and Immanuel Kant

Hume's Compatibilism reconciled freedom with the rational causal determinism of his age, the physical determinism of Newtonian classical mechanics.

Kant's great Critiques put limits on such reason, to make room for freedom, God, and immortality. He invented a transcendental noumenal world beyond the deterministic physics of the phenomenal world.


Reason and Free Will

Logically, Either Determinism is True or Indeterminism is True.

If Determinism is True, We Are Not Free.

If Indeterminism is True, our Will is Random, and we cannot be responsible for our actions. Logic and Reason find Free Will incompatible with both Determinism and Indeterminism.

This is the Standard Argument against Free Will.


Locke, Hegel, Darwin, and James

John Locke analyzed the complex concept of "free will" and found it unintelligible. "I think the question is not proper, whether the will be free, but whether a man be free."

G. W. F. Hegel synthesized indeterminism and determinism in the third step of his dialectical "Aufhebung" of free will.

Charles Darwin combined objective chance and lawful selection to explain biological evolution.

William James adapted Darwin in his two-stage model of "mental evolution."


Reconciling Indeterminism

David Hume reconciled freedom with the determinism of classical Newtonian physics . He found "no medium betwixt chance and necessity."

Arthur Eddington understood indeterminism and quantum physics , but he could find "no halfway house between randomness and correlated behavior."

I believe I can reconcile free will with the indeterminism of Heisenberg, by putting Kantian limits on it, and finding a Hegelian "third thing," one that combines indeterminism and determinism, while limiting and transcending both.


Aristotle, Epicurus, and a Third Thing

For Aristotle, human agency was beyond chance (τυχή) and necessity (ἀνάγκη). He called it "up to us" (ἐφ' ἡμῖν) and "in us" (ἐν ἡμῖν). It was a "third thing," a tertium quid.

Epicurus, known for the "swerve" of the atoms that breaks the deterministic causal chain, agreed there was a third thing…

"...some things happen of necessity (ἀνάγκη), others by chance (τύχη), others through our own agency (παρ’ ἡμᾶς).

"...necessity destroys responsibility and chance is uncertain; whereas our own actions are autonomous, and it is to them that praise and blame naturally attach. " (Letter to Menoeceus, §133)


Ontology of Alternative Possibilities

Alternative Possibilities may never be acted upon, and the idea of Actualism (only what actually happens is real) denies that they exist.

But alternative possibilities, once they are generated in the first stage of the two-stage process of first "free," then "will," become real - as information structures in the mind of the agent.

Barring memory loss, the information remains in the agent's mind, as "roads not taken," and as alternative possibilities when faced with similar circumstances in the future.


Information Flows

NoiseEntropy

ChaosDisorder

Boltzmann Entropy

InformationNegentropyComplexityOrder

Shannon Entropy

David Layzer showed how entropy and information can increase at the same time in the expanding universe. There are two info/entropy flows…

In any process, the positive entropy increase is always at least equal to, and generally orders of magnitude larger than, the negative entropy in any created information structures, to satisfy the second law.

Negative Entropy

OBJECTIVEGOOD?

OBJEC

TIVE

BAD?


Info Flows in Cosmology

Cosmic BackgroundDust and Gas?

Dark Matter?Dark Energy?

Boltzmann Entropy

Particles Quarks, Baryons Nuclei, Electrons Atoms, Molecules

GalaxiesStarsPlanets

Shannon EntropyNote that the stable quantum mechanical systems and self-gravitating systems have extremely long lifetimes, thanks to quantum stability.

Although quantum processes break the illusion of a chain of causation and pre-determinism, the most important contribution of quantum mechanics is to provide stability against quantum and thermal noise.

Negative Entropy


Sun-Earth Info Flows

Space

Boltzmann Entropy

Replicating MoleculesRNA-lifeDNA-life

Shannon Entropy

Most of the negative entropy flow from the sun is wasted, passing the earth and lost to outer space.

A tiny fraction is captured by the earth, keeping us at a temperature comfortable for life forms to evolve.

Negative Entropy


Info Flows in Biology

EnvironmentNight Sky

Boltzmann Entropy

ProtozoansPlants AnimalsHumans

Shannon Entropy

Every biological structure is a quantum mechanical structure.

DNA has maintained its stable information structure over billions of years in the constant presence of noise.

Negative Entropy


Info Flows in Humans

100W Heat LossCell death

ExcrementHuman Deaths

Culture Loss

Boltzmann Entropy

Body MaintenanceAnd GrowthMind - Learning

Knowledge Transfer PublicationsHuman Artifacts

Shannon Entropy

The stable information content of a human being survives many changes in the material content of the body. Only with death does human information (spirit, soul) dissipate - unless it is saved somewhere.

The total mental information in humans is orders of magnitude less than the information content and information processing rate of the body.

Negative Entropy


Information Creation in RBCs

100 million red blood cells die each secondx300 million hemoglobin proteins in each RBCx100s of amino acids in each hemoglobin =100,000 terabytes of information per second Every time a tRNA adds a

new amino acid to the growing polypeptide chain it is a quantum event!

Hemoglobin protein


Standard Argument against Free Will

To understand Kane's work, consider this argument.

Logically, Either Determinism is True or Indeterminism is True.

If Determinism is True, We Are Not Free.

If Indeterminism is True, our Will is Random, and we cannot be responsible for our actions.


Kane's Responsibility Either Way

Kane' Plural Rational Control cleverly develops cases where the agent is responsible, even if a decision is made at random.

In these cases the agent is agonizing over "torn decisions," and has excellent reasons for doing more than one thing.

This partially overcomes the problem of luck and the standard argument that indeterminism undermines responsibility.

But more is needed.


Chance Is Not Merely Epistemic

For centuries, the existence of regular statistical laws – the normal distribution of errors, etc. – suggested to many philosophers, scientists, and mathematicians that chance has an underlying determinism, that chance is only the result of human ignorance. This is a serious mistake.

Quantum statistics underlie regularities in classical statistics.

Quantum indeterminacy is the source of real ontological chance in the universe. Not surprisingly, the statistical distribution of quantum events matches that of the classical normal distributions (like the bell curve).

Documents

Department of Astronomy Harvard University