Determined: A Science of Life without Free Will

by Robert M. Sapolsky

This is the 80:20 rule—approximately 80 percent of interactions occur among approximately 20 percent of the population. In the commercial world, it’s sardonically stated as 80 percent of complaints come from 20 percent of the customers. Eighty percent of crime is caused by 20 percent of the criminals. Eighty percent of the company’s work is due to the efforts of 20 percent of the employees. In the early days of the pandemic, a large majority of COVID-19 infections were caused by the small subset of infected super-spreaders.[29]

Thus, in my view, emergent complexity, while being immeasurably cool, is nonetheless not where free will exists, for three reasons: Because of the lessons of chaoticism—you can’t just follow convention and say that two things are the same, when they are different, and in a way that matters, regardless of how seemingly minuscule that difference; unpredictable doesn’t mean undetermined. Even if a system is emergent, that doesn’t mean it can choose to do whatever it wants; it is still made up of and constrained by its constituent parts, with all their mortal limits and foibles. Emergent systems can’t make the bricks that built them stop being brick-ish.[*],[11]

People (e.g., Robert Brown, in 1827) had long noted the phenomenon, but it wasn’t until the last century that random (aka “stochastic”) movement was identified to occur among particles suspended in a fluid or gas. Tiny particles oscillate and vibrate as a result of being hit randomly by photons of light, which transfer energy to the particle, producing the vibratory phenomenon of kinetic energy. Which causes particles to bump into each other randomly. Which causes them to bump into other particles. Everything moving randomly, the unpredictability of the three-body problem on steroids.

I like teaching one example of Brownian motion, because it undermines myths of how genes determine everything interesting in living systems. Take a fertilized egg. When it divides in two, there is random Brownian splitting of the stuff floating around inside, such as thousands of those powerhouses-of-the-cell mitochondria—it’s never an exact 50:50 split, let alone the same split each time. Meaning those two cells already differ in their power-generating capacity. Same for vast numbers of copies of proteins called transcription factors, which turn genes on or off; the uneven split of transcription factors when the cell divides means the two cells will differ in their gene regulation. And with each subsequent cell division, randomness plays that role in the production of all those cells that eventually constitute you.[*],[2]

Consider some organism—say, a fish—looking for food. How does it find food most efficiently? If food is plentiful, the fish forages in little forays anchored around this place of easy eating.[*] But if food is diffuse and sparse, the most efficient way to bump into some is to switch to a random, Brownian foraging pattern called a “Levy walk.” So if you’re the only thing worth eating in the middle of the ocean, the predator that grabs you will probably have gotten there by a Levy walk.

To summarize, the world is filled with instances of indeterministic Brownian motion, with various biological phenomena having evolved to optimally exploit versions of this randomness. Are we talking free will here?[*] Before addressing this question, time to face the inevitable and tackle the mother of all theories.[4]

Next weirdness.[*] Two particles (say, two electrons in different shells of an atom) can become “entangled,” where their properties (such as their direction of spin) are linked and perfectly correlated. The correlation is always negative—if one electron spins in one direction, its coupled partner spins the opposite way. Fred Astaire steps forward with his left leg; Ginger Rogers steps back with her right. But it’s stranger than that. For starters, the two electrons don’t have to be in the same atom. They can be a few atoms apart. Okay, sure. Or, it turns out, they can be even farther apart. The current record is particles nearly nine hundred miles apart, at two ground stations linked by a quantum satellite.[*] Moreover, if you alter the property of one particle, the other changes as well, implying a causality that isn’t local. There is no theoretical limit for how far apart entangled particles can be. An electron in the Crab Nebula in the constellation Taurus can be entangled with an electron in the piece of broccoli stuck between your incisors. And as the strangest feature, when the state of one particle is altered, the complementary change in the other occurs instantaneously[*]—meaning that the broccoli and the Crab Nebula are influencing each other faster than the speed of light.[7]

When we argue about whether our behavior is the product of our agency, we’re not interested in random behavior, why there might have been that one time in Stockholm where Mother Teresa pulled a knife on some guy and stole his wallet. We’re interested in the consistency of behavior that constitutes our moral character. And in the consistent ways in which we try to reconcile our multifaceted inconsistencies.[*] We’re trying to understand how Martin Luther would stick to his guns and say, “Here I stand, I can do no other,” when ordered to renounce his views by ecumenical thugs who burned people at the stake as a hobby. We’re trying to understand that lost-cause person who is trying to straighten out their life yet makes self-destructive, impulsive decisions again and again. It’s why funerals so often include a eulogy from that person’s oldest friend, a historical witness to consistency: “Even when we were in grade school, she already was the sort of person who . . .”

The model of decision making I am proposing has the following feature: when we are faced with an important decision, a consideration-generator whose output is to some degree undetermined, produces a series of considerations, some of which may of course be immediately rejected as irrelevant by the agent (consciously or unconsciously). Those considerations that are selected by the agent as having a more than negligible bearing on the decision then figure in a reasoning process, and if the agent is in the main reasonable, those considerations ultimately serve as predictors and explicators of the agent’s final decision.[23]

Such a harnessing scenario has at least three limitations, of increasing significance: —A child has fallen into an icy river, and your consideration generator produces three possibilities to choose among: leap in and save the child; shout for help; pretend you didn’t see and scurry away. Choose. But since we’re dealing with quantum indeterminacy, what if the first three possibilities are: tango in the absence of a partner; confess to cheating on your taxes; make squawking sounds while jumping backward like the dolphins at Sea World? Perfectly plausible, if superpositioned electron waves are the wellsprings from which your moral decisions flow. —To avoid having only tangoing, confessing, and dolphining as options, determine that you need to indeterminately generate every random possibility. But now you have to spend a lifetime evaluating and comparing each before choosing which is best. You need to have an impossibly efficient search algorithm.[*],[25] —So, phew, generate enough options so that they aren’t all silly, figure out how to efficiently evaluate them all, and then use your criteria to filter out all but the winner. But where does that filter, reflecting your values, ethics, and character, come from? It’s chapter 3. And where does intent come from? How is it that one person’s filter filters out every random possibility other than “Rob the bank,” while another’s goes for “Wish the bank teller a good day”? And where do the values and criteria come from in even first ...

When people are made to feel helpless and with less agency by being stymied by an unsolvable puzzle, the size of their early readiness potentials decreases. And when people are prompted to believe less in free will, the same occurs, with less belief predicting a greater blunting of the wave (without changing the size of the subsequent wave in the motor cortex itself)—people seem to not be trying as hard, focusing as hard on the task.[5]

A series of studies initiated by behavioral economist Katherine Vohs of the University of Minnesota show that free-will skeptics become more antisocial in their behaviors. In experiments, they are more likely to cheat on a test and to take more than their fair share of money from a common pot. They become less likely to help a stranger in need and more aggressive (after being rebuffed by someone, the subject gets to take revenge by determining how much hot sauce the person will have to consume—make someone a free-will skeptic and they nearly double the amount of retributive hot sauce). Less free-will belief and subjects feel less grateful to someone who has done them a favor—why feel gratitude for an act that was someone’s mere biological imperative? And just in case it seems like these skeptics are now having too much nihilistic fun by getting to take revenge with a dish served spicily, the manipulation also makes people feel less meaning in their lives and less of a sense of belonging to other humans. Moreover, lessened free-will belief leads to people feeling like they have less self-knowledge, and to feeling alienated from their “true selves” when making a moral decision.

Thus, undermine someone’s belief in free will and they feel less of a sense of agency, meaning, or self-knowledge, less gratitude for other people’s kindness. And most important for our purposes, they become less ethical in their behavior, less helpful, and more aggressive. Burn this book before anyone else stumbles upon it and has their moral compass unmoored.

Collectively, these studies show that religious primes bring out the best in religious people, making them more charitable, generous, and honest, more resistant to temptation, and more capable of exerting self-control.

Now we’re getting somewhere. When religious people are not thinking about their religious principles, they sink into the same immoral muck as atheists. But remind them of what really matters, and the halo comes out. Two big complications: The first is that in a lot of these studies, implicit religious primes make atheists more prosocial as well. After all, you don’t have to be a Christian to decide that the Sermon on the Mount has good parts. But as a more informative complication, while prosociality in religious people is boosted by religious primes, prosociality in atheists is boosted just as much by the right kinds of secular primes. “I’d better be good or else I’ll get into trouble” can certainly be primed by “alij” or “eocpli.” Prosociality in atheists is also prompted by loftier secular concepts, like “civic,” “duty,” “liberty,” and “equality.”[*],[25]

The highly religious tend to view good works more in a personal, private context, helping to explain why religious Americans donate more of their income to charity than do the secular. In contrast, atheists are more likely to view good works as a collective responsibility, helping to explain why they are the ones who are more likely to support candidates advocating wealth redistribution to decrease inequality. Thus, if you’re trying to decide who is more likely to run amok with antisocial behaviors, atheists will look bad if the question is “How much of your money would you give to charity for the poor?” But if the question is “How much of your money would you pay in higher taxes for more social services for the poor?” you’ll reach a different conclusion.[27]

The preceding sections suggest that deciding that there is no omnipotent being to punish transgressions doesn’t send atheists into a downward moral spiral. It should be noted, however, that a huge percentage of the research discussed has been with American subjects, from a country where only roughly 5 percent of people say they are atheists. We saw that prosociality can even be enhanced in atheists by religious primes. Maybe the relative morality of atheists is due to being surrounded by the morality of all those theists rubbing off on them. What would happen if most people became atheist or irreligious—what sort of society would they construct, when everyone is freed from being nice due to the fear of God?

Even after controlling for factors like self-reporting or demographic correlates of religiosity, and after considering broader definitions of prosociality, religious people still come through as being more prosocial than atheists in some experimental as well as real-world settings. Which leads us to a really crucial point: religious prosociality is mostly about religious people being nice to people like themselves. It’s mostly in-group.

How about when it comes to out-group members? In those circumstances, it is atheists who are more prosocial, including more accepting of and extending protection to Thems. Moreover, religious primes can make religious people more prejudiced against out-group members, including increasing vengefulness and willingness to punish their transgressions.

The answer is that we don’t change our minds. Our minds, which are the end products of all the biological moments that came before, are changed by circumstances around us.

Take one lone heroic rat that, for some reason, can save the world from disaster by developing a conditioned eyeblink response. And he screws up, doesn’t do it, lets the world down. Afterward, everyone is pissed at the rat, blaming him for not conditioning. To which he can say, “It’s not my fault—I didn’t get conditioned because, one second before, my interpositus nucleus wasn’t as responsive to the conditioned stimulus; because a few hours before, my stress hormone levels were elevated, which guaranteed that the interpositus would be particularly resistant to conditioning; because back in my childhood, my mother was taken from me, and this changed gene regulation in the interpositus, permanently increasing levels of a hormone receptor there; because back millions of years ago, my species evolved to be highly dependent on maternal care after birth, and the genes needed to make lifelong changes in circuitry if the mother is absent.” A change in behavior due to specific changes that can be identified in a circuit, arising from circumstances a second, an hour, a lifetime, an evolutionary epoch earlier over which the organism had no control. No rodential moral responsibility involved, no grounds for everyone blaming the rat.

We saw that something as simple as eyeblink conditioning reflects a nervous system that has been sculpted by all that came before it (e.g., early maternal experience). The acquisition, consolidation, and extinction[*] of the conditioned fear of something neutral like a tone reflects the organism’s history even more. Extinction will occur faster if, in the seconds before, there are high levels in the amygdala of endocannabinoids (whose receptor also binds THC, the most active component of cannabis)—this makes it easier to stop being afraid of something. The amygdala becomes less likely to store away a conditioned fear response as a stable memory if, in the previous hours, the individual has taken an SSRI antidepressant like Prozac (which makes people ruminate less about negative thoughts). The amygdala will be less active and harder to condition if, in the days before, it was exposed to high circulating levels of oxytocin, which helps explain how oxytocin can promote trust.

2001 paper describes one such case, of a woman whose rare, intractable epilepsy produced seizures virtually daily that were associated with outbursts of agitated aggression. She had been arrested thirty-two times for such violent incidents; the severity of violence escalated, culminating in a murder. The seizure focus was near the amygdala, and after surgical removal of that part of her temporal lobe, both the seizures and the aggressive outbursts stopped.[4]

A great demonstration of this is the fact that if someone has the disease, their identical twin, who shares all their genes, has a 50 percent chance of having it as well (versus the usual 1–2 percent risk in the general population). The genetics of schizophrenia, however, are not about a single gene that has gone awry (as compared with classic single-gene disorders such as cystic fibrosis, Huntington’s disease, or sickle cell anemia). Instead, it arises from an unlucky combination of the variants of an array of genes, many of which are related to neurotransmission and brain development.[*] However, the collection of genes does not cause schizophrenia but, instead, increases the risk for it. This is implicit in flipping the finding just mentioned on its head—if someone has the disease, their identical twin has a 50 percent chance of not having it. In a classic gene/environment interaction, getting the disease basically requires a combination of the genetic vulnerability plus a stressful environment. What sort of stress? During fetal life, disease risk many years later is raised by prenatal malnutrition (for example, the Dutch Hunger Winter famine of 1944 greatly boosted the incidence of schizophrenia among individuals who had been fetuses at that time), exposure to any of a number of viruses by way of maternal infection, placental bleeding, maternal diabetes, or infection with the protozoan parasite Toxoplasma gondii.[*]

Aberrant salience is thought to also contribute to another defining feature of the disease, namely the hallucinations. Most people have an internal voice in our heads, narrating events, reminding us of things, intruding with unrelated thoughts. Have a random burst of dopamine along with one of those, and it becomes marked with so much salience, so much presence, that you perceive it, respond to it as an actual voice. Most schizophrenic hallucinations are auditory, reflecting how much of our thinking is verbal. And as a truly remarkable exception that proves the rule, there have been reports of congenitally deaf individuals with schizophrenia whose hallucinations are in American Sign Language (where some hallucinate a pair of disembodied hands signing to them, or being signed to by God).[*],[23]

In the early 1990s, about a third of the soldiers deployed in the first Gulf War complained of being “never quite right again,” with a constellation of symptoms—exhaustion, chronic unexplained pain, cognitive impairments. “Gulf War syndrome” was generally viewed as being some sort of psychological disorder, i.e., not for real, a marker of psychologically weak, self-indulgent veterans. And then science trickled in. Soldiers had been administered a heavy-duty class of drugs related to pesticides as protection against the nerve gas that Saddam Hussein was expected to use. While these drugs could readily explain the neurological features of Gulf War syndrome, this was discounted—careful research in the run-up to the war had identified what doses could be given safely, would not damage brain function. But then it turned out that the drugs became more damaging to the brain during stress, something that was not considered beforehand. One of the mechanisms implicated was that stress—in this case, body heat generated by carrying eighty pounds of gear in 120-degree desert weather, coupled with basic combat terror—could open up the blood-brain barrier, increasing the amount of drug getting into the brain.

sybarite

The ubiquity of cooperation among social species raises a ubiquitous problem. Sure, it’s great when everyone cooperates for the greater good, but it’s even better when everyone else does that while you mooch off them. This is the problem of cheating. A lioness conveniently lags behind the others in a dangerous hunt; a bat doesn’t feed the others’ kids but freeloads on their cooperation; a baboon stabs his coalitional partner in the back. Two separate colonies of genetically identical social amoebas merge to form a multicellular structure called a fruiting body, which consists of a stalk, which gives stability, and a cap. Only the amoebas in the cap reproduce, and cooperation consists of each colony equally sharing the bummer of being nonreproducing stalk cells; instead, different strains will try to cheat, exploiting the other colony by preferentially hogging seats on the cap. Even mitochondria and stretches of DNA cheat in cooperative ventures.[*],[15] And sure as the day is long, the ubiquity of cheating drove the evolution of countermeasures to detect and punish it. Chimps that fail to support an ally in a fight are pummeled afterward. Wrens that don’t feed the nestlings of the dominant breeding pair are attacked. Naked mole rat queens are aggressive toward workers that are slacking. In the mutualism where reefer fish are cleaned by wrasse fish that get to eat the parasites they harvest, some wrasses cheat to get an even better meal by taking a bite out of the reefer fi...

The right kind of punishment at the right time matters for enhancing cooperation.

player double-crosses you; at that point, do it back in the next round. If they continue cheating, continue punishing them back, but if they go back to cooperating, you resume the same in the next round. A strategy that has clear rules, that starts with cooperation, that is proportionately punitive against cheaters, and that can forgive.

Cool cross-cultural research shows that small, traditional cultures—say, hunter-gatherers or subsistence farmers—don’t carry out third-party punishment (either in real life or when playing economic games). They fully understand when cheating occurs but just don’t bother. Explanation: everyone knows each other and what they’re up to, so you don’t need fancy third-party enforcement to rein in antisocial behavior. Supporting this, the larger the society, the more formalized the third-party policing. In addition, fourth-party punishing of third-party cheating works best when there are only a small number of third-party enforcers—think of the chaos if, instead of police, things ran solely on citizens’ arrests.[20]

A way to lower the cost of punishing involves reputation, an incredibly reliable means of influencing behavior. In tests of game theory, cooperation is boosted if people know your history of play (i.e., open-book play that produces a shadow of the future); be known as a free rider, and others will start off not trusting you or refusing to play with you. This occurs among hunter-gatherers, who spend a huge amount of time gossiping about, among other things, who has cheated by, say, not sharing meat; get a reputation for that and you’re ostracized, which can be life-threatening. In contrast, the costs of third-party punishment are reduced because your reputation is enhanced and people trust you more; if you’re already viewed as socially dominant, being a third-party punisher makes you seem more formidable and likable.[23]

Why do animals mate, expending effort and calories, often risking their lives? Distal explanation: because it allows you to leave copies of your genes in the next generation. Proximal explanation: it feels good. Why punish cheaters when it’s costly? The distal explanation is what we’ve been discussing—because reliably and collectively sharing the costs benefits everyone. But it is when we look for a proximal explanation that we see how it’s going to be so damn hard to get people to proclaim the lack of free will and just quarantine the dangerous. Why punish cheaters when it’s costly? Proximally, because we like to punish wrongdoers. It feels great.

raucous

Here’s a thoroughly elegant study, carried out by German psychologist Tania Singer. Subjects were either six-year-old kids or chimps. One of the researchers comes into the room and either does something nice to the kid/chimp—offering some desirable food—or does something mean—teasing them by starting to give the food and then snatching it away. The researcher leaves and then enters an adjacent room, visible to the subject through an observation window. Someone sneaks up behind the researcher and—whoa!—seemingly starts hitting them over the head with a stick, with the researcher crying out in pain. After ten seconds, the assailant drags the researcher to an adjacent room and then resumes the hitting. The kid/chimp can go into their own adjacent room with another window, giving them the opportunity to watch. Do they move to do this? If the researcher being pummeled had been nice to them, only 18 percent moved to see the rest; if the researcher had been mean, 50 percent leaped at the opportunity. Both kids and chimps were particularly interested in seeing someone who was mean to them get punished.[26]

The same unsettling conclusion comes from neuroimaging studies. If someone makes you an unfair offer in the Ultimatum Game, your insula, anterior cingulate cortex, and amygdala activate, a picture of disgust, pain, and anger. The lowball offer puts you at a split in the road. If it’s a single-round game, punish retributively or be purely logical and accept the offer that is better than nothing? The more activation of your insula and amygdala, and the more pissed off you report being by inequity in general, the more likely you are to reject the offer. This retributive irrationality is all about emotion—if people believe they are rejecting an unfair offer from a human rather than a computer, there is also activation of that emotional vmPFC; making a similar point, men with higher testosterone levels are more likely to reject such offers.[28]

So on a neurobiological level, second-party punishers are about disgust, anger, and pain, whereas third-party punishers have the same plus the perspective taking needed to view someone else’s misfortune as akin to your own. But then there is the crucial additional finding in all these cases: retributive punishing in any of these guises also activates the dopamine circuitry involved in reward (the ventral tegmentum and nucleus accumbens). Activation by punishment of the brain region goosed by the likes of orgasm or cocaine. It feels good.[30]

Being altruistic can feel good—it decreases pain in cancer patients, blunts the activation of neural pain pathways in response to shock. It even literally gives you a warm glow (such that people estimate ambient temperatures as being higher after an altruistic act). Nice. But being able to righteously punish evildoers feels really good. But as will be seen in a bit, even that can be tamed.[31]

paean

A whole field of psychology explores terror management theory, trying to make sense of the hodgepodge of coping mechanisms we resort to when facing the inevitability and unpredictability of death. As we know, those responses cover the range of humans at our best and worst—becoming closer to your intimates, identifying more with your cultural values (whether humanitarian or fascist in nature), making the world a better place, deciding to live well as the best revenge. And by now, in our age of existential crisis, the terror we feel when shadowed by death has a kid sibling in our terror when shadowed by meaninglessness. Shadowed by our being biological machines wobbling on top of turtles that go all the way down. We are not captains of our ships; our ships never had captains.[2] Fuck. That really blows.

Humans “descended from the apes! Let us hope it is not true, but if it is, let us pray that it will not become generally known,” said the wife of an Anglican bishop in 1860, when told about Darwin’s novel theory of evolution.[*] One hundred fifty-six years later, Stephen Cave titled a much-discussed June 2016 article in The Atlantic “There’s No Such Thing as Free Will . . . but We’re Better Off Believing in It Anyway.”[*]

And thus, perhaps, “we’re better off believing in it anyway.” Truth doesn’t always set you free; truth, mental health, and well-being have a complex relationship, something explored in an extensive literature on the psychology of stress.

Expose a test subject to a series of unpredictable shocks, and she will activate a stress response. If you warn her ten seconds before each shock that it is coming, the stress response is lessened, as truth girds predictability, giving time to prepare a coping response. Give a warning one second before each shock, and there’s too little time for an effect. But give a warning one minute before and the stress response is worsened, as that minute stretches into feeling a year’s worth of anticipatory dread. Thus, truthful predictive information can lessen, worsen, or have no effect on psychological stress, depending on the circumstances.[3]

Researchers have explored another facet of our complex relationship with truth. If someone’s actions have produced a mildly adverse outcome, truthfully emphasizing the control he had—“Think how much worse things could have been, good thing you had control”—blunts his stress response. But if someone’s actions have produced a disastrous outcome, untruthfully emphasizing the opposite—“No one could have stopped the car in time, the way that child darted out”—can be deeply humane. The truth can even be life-threatening. Someone teetering on the edge of death in an ER, 90 percent of their body covered in third-degree burns, gathers their strength to ask in a whisper whether the rest of their family is okay. And most medical professionals would be mighty torn about telling the person the shattering truth. As some evolutionary biologists ...

According to Daniel Dennett, if there were no belief in free will, “there would be no rights, no recourse to authority to protect against fraud, theft, rape, murder. In short, no morality.

As we’ve seen, rejection of free will doesn’t doom you to break bad, not if you’ve been educated about the roots of where our behavior comes from. Trouble is, that requires education. And even that doesn’t guarantee a good, moral outcome.

Rejecting free will has an additional downside. If there’s no free will, you don’t deserve praise for your accomplishments, you haven’t earned or are entitled to anything.

In the short story “What’s Expected of Us,” Ted Chiang takes a cue from Libet, writing about a gizmo called the Predictor, with a button and a light. Whenever you press the button, the light goes on a second before. No matter what you do, no matter how much you try not to think about pressing the button, strategize to sneak up on it, the light comes on a second before you press the button. In the moment between the light coming on and your supposedly freely choosing to press the button, your future action is already a determined past. The result? People are hollowed out. “Some people, realizing that their choices don’t matter, refuse to make any choices at all. Like a legion of Bartleby the Scriveners, they no longer engage in spontaneous action. Eventually, a third of those who play with a Predictor must be hospitalized because they won’t feed themselves. The end state is akinetic mutism, a kind of waking coma.”[7]

Maybe you’re deflated by the realization that part of your success in life is due to the fact that your face has appealing features. Or that your praiseworthy self-discipline has much to do with how your cortex was constructed when you were a fetus. That someone loves you because of, say, how their oxytocin receptors work. That you and the other machines don’t have meaning.

And all the while, chaoticism teaches us that “being normal” is an impossibility, that it ultimately just means that you have the same sorts of abnormalities that are accepted as out of our control that everyone else has. Hey, it’s normal that you can’t cause objects to levitate.