Your decision-making ability is a superpower physics can’t explain
In a universe that unthinkingly follows the rules, human agency is an anomaly. Can physics ever make sense of our power to change the physical world at will?
Richard Webb
Physics 12 February 2020
By deck of cards artwork
I’VE been thinking about getting a puppy. You know, for a bit of companionship, something to motivate on grey days when spirit and flesh are weak.

I even went to a stray dogs’ home, because that seemed the right thing to do. There was a lovely one there, with beautiful, mischievous eyes. She reminded me of a mutt we had when I was a kid, called Whiskey. I bottled it in the end, though. Did I really have the time to give her the love and attention she deserved?

Whims, memories, hopes, judgements, morals, qualms – all coming together to influence decisions. It is hard enough for us to understand how we reach them. For fundamental physicists, it is a complete mystery. That is because our decision-making ability is a not-so-secret superpower to alter the physical world, changing its evolution apparently at will – something no physical law yet devised can explain.

“We act, we decide, we initiate actions,” says Carlo Rovelli at Aix-Marseille University in France. “How can we insert this agency into the general picture of nature?”

Rovelli and others have undertaken to find out. Their journey has led them into the depths of the human mind and its relationship with physical reality, throwing up surprising and profound connections: to the mysteries of entropy and flowing time, to reality and consciousness, and to the nature of physical law itself. Get to grips with what underlies our everyday acts, and we could be on the way to a deeper, all-inclusive understanding of both the cosmos and our place in it.

At its simplest, agency is relatively easy to define. “It is just the notion that certain systems in the world have intentional states, desires to bring stuff about,” says philosopher Eleanor Knox at King’s College London. “We’re clearly systems like that.”

“Physics aims to characterise everything through cast-iron laws with predictive power”

The arguments start with what else is too. “Quantum fields don’t have any agency. Atoms don’t, do bacteria?” asks physicist Sean Carroll at the California Institute of Technology. “I don’t know, but human beings do. Somewhere along that continuum it sneaked in.”

Philosophers and theologians have been poring over that and agency’s relationships to other thorny concepts, such as consciousness and free will, for millennia (see “Life, agency and everything”). But it is since humans started to do physics that agency has taken on particularly puzzling proportions.

The aim of physics is to characterise the interaction and evolution of reality’s elements through cast-iron mathematical laws with predictive power. That mission is far from complete, but it has been stunningly successful so far. From Isaac Newton’s laws of motion and gravitation to Albert Einstein’s relativity and the enigmatic edifice of quantum theory, we now have laws explaining everything from how apples fall to how biological and chemical processes unfold to the origin and fate of the universe.

Take an apple’s sudden detachment from a tree. It is caused by a complex series of biochemical processes, plus the whims of wind and weather, all explicable by the laws of physics. Given enough computational power, in theory we could trace the chain of causation back through Earth and cosmic history to the moment the carbon atoms within the apple were first manufactured in a supernova – and beyond that almost to the big bang and time’s beginning.

The thing is, that apple was always going to fall; it can’t choose to jump or decide not to fall. Agency is different. From our perspective at least, our decision-making is discretionary; there is none of the inevitability that marks out physical laws. And the chains of causation behind our actions are very complex, running through psychological influences brewed by nature, culture and nurture.

melting ice cream
Ice creams don’t choose to melt – but we choose to buy one
Hello World/Getty Images
Our current physical laws don’t try to explain this. “We think of ourselves as coming from outside the causal order and somehow intervening in it, making things happen,” says philosopher Jenann Ismael at Columbia University in New York. That strange detached position is maintained – albeit with an additional twist – even in the weird world of quantum theory, our most advanced underlying theory of material reality (see “No ‘I’ in collapse”).

The god-like status we accord ourselves is highly suspicious to many physicists. “If I’m saying that something doesn’t boil down to the laws of physics, then I’m basically positing something supernatural, that’s outside natural laws,” says Matt Leifer at Chapman University in California.

“Our decision-making has none of the inevitability that marks out physical laws”

That has been a popular way out for natural philosophers over the years, in various forms of mind-body dualism: the idea that the mental and physical realms are separate, and the rules of one don’t apply to the other. But that hardly seems a tenable position within modern science. “Being a full-on dualist is quite hard because it does look like, for instance, when I put lots of serotonin in your brain, your mental states change,” says Knox. “The question is how you think that could work if you think there’s two kinds of separate stuff.”

“If there’s evidence we should carve out a different realm for organic things or people or whatever, then by all means,” says Carroll. “But I’m made of atoms, my laws of physics purport to explain atoms, and it would seem by far the most likely hypothesis that the laws of physics explain me.”

The central conundrum becomes what sort of physical laws can unify two very different, conflicting views. “We see agents that make choices and exert a causal influence on what happens in the world, and then science comes along and says, ‘You’re actually a bunch of particles or atoms and you’re just obeying differential equations’,” says Carroll. “What we want to figure out is how those things can both be true at the same time.”

Agency is bound up with several other fundamental concepts that philosophers, physicists and others struggle to get to grips with. That starts with …

LIFE. In the early 1990s, an advisory panel to NASA defined life as “a self-sustaining chemical system capable of Darwinian evolution”. Others dissent. Typical defining characteristics include being born and dying, metabolising, growing and reproducing. Some life forms are known to exhibit more complex characteristics, including…

INTELLIGENCE. Broadly, the ability to learn stuff and solve problems. Artificial machine-learning systems can do those things too, so life probably isn’t a prerequisite. But AI is still far away from the “general intelligence” of humans and some animals that allows knowledge gained in one area to be applied elsewhere. This is possibly bound up with…

CONSCIOUSNESS. A perennial head-scratcher, consciousness starts with the observation that some things are aware they exist. Some conscious beings argue that explaining this state in purely physical terms will always fall foul of the “hard problem“: that certain aspects of it, such as individual experiences of colour or pain or love, can only ever be felt inwardly. The difficulty of defining consciousness satisfactorily leads some physicists to prefer…

AGENCY. This is at least expressed outwardly, in actions that change the external world. Human-type agency is arrived at intentionally, through internal decision-making processes, which throws up thorny questions of…

FREE WILL. Often seen as the opposite of determinism (everything is pre-determined), free will in its strongest sense, the complete freedom to act independently of any causal influences at all, almost undoubtedly doesn’t exist. The sort of free will we experience – agency that can be free of immediate physical causation – isn’t necessarily incompatible with either the deterministic laws of classical physics or the indeterminate probabilistic laws of quantum mechanics. The question is how agency fits in with those laws – and whether it should.

For Carroll and many others, the answer lies not in mysticism, but in emergence. This is the idea that behaviours and properties that are inscrutable when you look at single components of a complex system pop into existence when you view things as a whole. There are plenty of precedents. The temperature or density of a gas, for example, doesn’t mean much at the level of single molecules. Look at all the molecules of the gas together, however, and they are measurable quantities that explain physical change: how temperature differences cause heat flows, for example, or how a gas pushes a piston when compressed.

The catch is that human brains are phenomenally complex. A fully grown one contains nearly 100 billion interconnected neurons, and we are still far from establishing how they produce the felt states of our conscious world, let alone how those states interface with the wider world. “It’s one thing to say I can explain the temperature and density of the air by hypothesising that it is made of molecules bumping into one another,” says Carroll. “It seems quite a more dramatic project to say I can explain mind and choices and consciousness as emerging out of atoms and molecules bumping into one another.”

But just as we don’t need to know how every molecule in a gas is moving to know its temperature, so agency might be understood by skating over its interior details and finding global quantities that correspond to measurable outcomes. “I want to somehow find the minimal intellectual tools for understanding how something can work that you would describe as an agent and still be compatible with physics,” says Rovelli.

A matter of time
We have at least one big clue where to dig: the way agency changes the future, but not the past. “We think of agency in a time-oriented way,” says Rovelli. “We do something and then something happens.” Most physical laws don’t work like this: the basic equations of classical and quantum physics run just as well backwards as forwards.

The only one-way street in physics is the inescapable rise of disorder, or entropy. This is encapsulated by the second law of thermodynamics, the empirical law that says ice creams melt, milk can’t be unspilled and that it is far easier to lose one sock than to unite a pair. “That’s one of the steps to clarify to unravel this problem,” says Rovelli. “How do entropy and the second law of thermodynamics come in?”

Rovelli, Carroll and others are attempting to find out by sketching the possible connections between agency and wider cosmic flows. Carroll admits that there is still a lot of groundwork to be laid. “How do you even translate concepts like ‘make a decision’ or ‘choice’ or ’cause something to happen’ into the language of statistical mechanics where you have things bumping into each other and probability distributions and stuff like that?” he asks.

Hidden away in these thermodynamic considerations is the concept of information. A century and a half ago, physicist James Clerk Maxwell imagined a tiny intelligent being that could gauge the speed of individual molecules in a gas to sort hotter ones from cooler ones, and so apparently decrease disorder and entropy in contravention of the second law.

“Our agency is time-oriented: we can change the future, but not the past”

Subsequent investigations showed that it was clever manipulation of information that gave “Maxwell’s demon” this apparently unphysical power. Much more recently, physicists including Paul Davies at Arizona State University have suggested that information manipulation might be what distinguishes living matter from inert matter.

It almost certainly has a bearing on agency. A curious physical fact is that we have access to detailed information about the past, in the form of our memories and other evidence, and also more limited, unreliable and corrupted information about the future, thanks to prediction from patterns in the past and present. Does agency in some way consist of an ability to harness and combine these different forms of information – and if so, can we develop a mathematical language to express their differences? Carroll is working on that. “I think that I have ideas and I’m on the right track for understanding,” he says. “But the i’s have not yet been dotted.”

Others are scratching around the same territory. “Agents are just creatures that can learn about the world and use what they learn to change and control it,” says Susanne Still at the University of Hawaii. “They follow rules of information acquisition, so we need to find out what those rules are.”

road sign
Every time we make a decision, we change the evolution of the universe
Maciej Bledowski/Alamy Stock Photo
Those rules must fulfil certain criteria. They must include some element of memory storage and recall: the ability to bank and process external information over time so as to project it onto the future. They must allow for feedback loops, so relevant new information can update information already held internally. There must also be a way of delineating an agent’s boundary, to distinguish actions triggered internally from those triggered externally.

“Agency is not just reflexes,” says Larissa Albantakis at the Center for Sleep and Consciousness at the University of Wisconsin-Madison. “If you’re only reacting to the environment, you’re not an agent, you’re just a system going through the motions.” The apparently deliberative quality of our agency sets it apart from, say, bacteria responding to chemical stimuli, or even frogs reflexively snapping at passing flies. “We collect influences from our past, we subject those influences to reflective process, we somehow extract things like hope and dreams and bring them to bear on behaviour, to mediate between the influences impinging on us,” says Ismael.

Together with her colleague Giulio Tononi, Albantakis has recently shown how systems with the same overall physical dynamics, but different internal ways of dealing with and distributing information, can develop different degrees of autonomous agency. Meanwhile, Still has found that information-acquiring machines built to retain memories and operate at maximum thermodynamic efficiency necessarily only retain new information with predictive power. “What emerges is an information bottleneck – a method that tells you to distil relevant, predictive bits and to throw away irrelevant clutter,” she says. Because thermodynamic efficiency gives a survival advantage, that might be a further lead on how something like agency first arose.

Many other researchers are working at different bits of the agency problem, but all admit it is a huge work in progress. “I recognise the largeness of the project and the smallness of our progress so far,” says Carroll.

Mind over matter?
And there might be a twist to the tale. In the end, it might turn out to be less about what physics can tell us about agency, and more about what agency can tell us about physics.

“The question of how much of the structure that I see around me is my concepts projected onto the world, and how much is the world projected onto me, is one of the deepest in the philosophy of mind,” says Knox. All we can say for certain about the laws of physics is that they make sense to us. Useful as their predictive power may be, we have no guarantee of their relationship to fundamental reality. Given these limitations, should we accept the starting premise that only they can provide answers?

Leifer for one is doubtful. “I don’t believe that physics is necessarily as fundamental as most of us have been led to believe,” he says. Physics has been so successful, he thinks, precisely because it has extracted the easy stuff – the bits of the world amenable to characterisation by regular, mathematical laws – and put them in a box marked “physics”. But that doesn’t mean everything fits in there.

Take an old chestnut that often comes up when people talk about conscious perceptions: colour. “Physicists have a definition of red: light of such and such a wavelength,” says Leifer. But they miss out the most essential aspect of a thing’s redness – how red we perceive it to be – purely because we have no way of coming up with a common standard. “And why should we expect physics to have anything to do with it?” asks Leifer. Agency might represent a similar conundrum, in which case we are fated to remain beyond the reach of the universe of physics we have invented.

“The whole idea of an external world evolving by regular laws might be an illusion”

There are even some indications that “invented” really is the right word. Theoretical physicist Markus Müller at the Austrian Academy of Sciences in Vienna has recently shown that physical laws, at least of the type that underlie quantum theory, could be brought into existence purely by modelling how agents combine information probabilistically to come to decisions. In that case, the whole idea of an external world evolving according to regular laws might be an illusion: agency is the only thing there really is.

A bridge too far? Maybe. Other answers are available. Perhaps they lie in some new understanding of how quantum effects play out in our brain, for example, or even more speculatively in the interplay of quantum theory and gravity, or other physics we haven’t even invented yet.

“Is it just a matter of building the bridges between different layers of scientific description, or understanding an entirely new phenomenon?” asks Carroll. “I think it’s the first answer, but I’m admitting there’s a question there.”

For all the uncertainties, asking these questions is a worthwhile enterprise, says Ismael. “It’s a really exciting development that physicists are trying to understand the human being and its place in nature,” she says. Finding answers will ultimately depend both on physicists’ calculating nous and philosophers’ clear conceptual analysis, she thinks. “This really is a place where physics and philosophy can fruitfully interact.”

Where might this take us? “In the end, success will be a naturalistic understanding of human beings that seems to answer to our own conception of who and what we are, in ways that support things that matter about us, like moral responsibility and our sense that when we’re making a decision, that decision is playing an indelible and pivotal role in what we do,” says Ismael.

There is a limit to how far that goes, of course. Physics is unlikely to give us any guidance as to whether we are climbing the right way up life’s many-branched decision tree. Should I have got that puppy? I truly can’t decide.

The role of observers in quantum mechanics is highly mysterious
Victor de Schwanberg/Science Photo Library
When it comes to the relationship between our power of agency and physical laws, quantum theory at first seems to send us spiralling into a new total perspective vortex.

The difficulties are best illustrated by Schrödinger’s cat, the notorious feline that exists both killed and not killed by the random quantum decay of a radioactive atom, until – in the standard telling at least – an observer determines which by looking in their kitty carrier.

Erwin Schrödinger was motivated to this absurdist thought experiment by a seemingly absurd reality. In quantum theory, objects are found not in definite states, but as probabilistic “wave functions” that allow the simultaneous existence of several different possible states. Make a measurement, however, and a single reality emerges. Quantum experiments back this up. Repeatedly measure the same quantum object and you won’t measure the same state each time. Instead, a pattern of results will slowly develop corresponding to the probabilities of the different states encoded in the wave function.

At face value, the process of “wave function collapse” to create a single defined reality brings an unwelcome new twist on agency: we as observers get to bring reality into being, but with no power over what we get. “Quantum mechanics doesn’t insert us into the causal chain,” says philosopher Jenann Ismael of Columbia University in New York. “It inserts uncontrollable events into the causal chain.”

Things get even weirder with recent, more complex versions of Schrödinger’s thought experiment in which two different agents simultaneously see contradictory versions of the same reality. So does agency break quantum theory?

Probably not. More plausibly, this is evidence that we need to work on our interpretations of quantum theory, and that versions that credit observers with creating reality don’t cut it. Many physicists don’t need convincing. “It’s plainly bananas to believe you’ll have no measurement results until the first human life evolved,” says Matt Leifer at Chapman University in California.

Others see in it vindication for their own interpretations. Sean Carroll at the California Institute of Technology, for example, is one of many adherents of the many-worlds interpretation of quantum mechanics. This says that when we make a quantum measurement, we don’t decide anything, but are merely conveyed with the result of our measurement into one of many parallel realities corresponding to all the possibilities encoded in the wave function. “Despite the name, many worlds has the simplest, smallest, most compact fundamental picture of reality,” says Carroll. “It’s just a wave function obeying an equation.”

The consequences might seem dramatic, with an infinity of parallel universes popping into existence at every fork in the road, but at least it would mean there is no particular quantum mystery to agency. “Attaching any special notion of agency to the collapse of the wave function makes no sense,” says Carroll. “But things like how do I choose what clothes to wear? That’s a perfectly reasonable question where agency is obviously involved.”

Jim Al-Khalili will shine a light on the most profound insights revealed by modern physics at an evening lecture in London on 11 March