Neurological Correlates Allow Us to Predict Human Behavior
Neurological Correlates Allow Us to Predict Human Behavior

Neurological Correlates Allow Us to Predict Human Behavior

A combination of factors, from oxytocin release as an indicator of emotional investment to cortisol and other hormones that correlate with attention, can forecast what people will do after an experience.

Paul J. Zak
Oct 1, 2020

ABOVE: © ISTOCK.COM, MAKSIM TKACHENKO

When he asks if I need help, I am mortified. At 38,000 feet, the stranger next to me believes I’m having a crisis. Maybe I am. I’m crying uncontrollably as I watch the scene unfolding before me on the seven-inch screen. Damn that Clint Eastwood. 

Movies, songs, and photographs can bring us to tears of joy or sadness. Why did Million Dollar Baby reduce me to a quivering mess while other people enjoy it without the emotional overflow? My lab has spent much of the last 20 years measuring brain activity to predict when an experience will induce emotional reactions. Our research established the neurochemical oxytocin as a key signal that the brain values an experience and thereby affects people’s decisions, and a vast literature now describes oxytocin’s role in motivating prosocial behaviors such as trustworthiness, generosity, and charitable giving. Recently, we found that combining measures of oxytocin with other neurochemicals allows us to build models that accurately predict what individuals and populations of people will do—such as whether they will be invested enough in Hilary Swank’s character to cry when she dies. 

Neural measures are valuable because people have difficulty articulating the motivations for their actions. As a result, survey-based research can lead organizations that supply goods and services to waste time and money offering people the wrong things. To improve market assessments, my team and I have worked to identify what neurological correlates are able to forecast human behavior. We even launched a startup and developed a software platform so that companies can use neural responses to predict which TV shows and movies will be hits and which songs will reach Billboard’s number one, as well as to identify learners who have effectively understood education and training and to drive up productivity after virtual meetings. While scientists are just beginning to use neurological measurements to predict what people will do, the technology holds great promise to improve people’s lives.

Oxytocin: A biomarker of trust

Since its discovery by English pharmacologist and physiologist Henry Dale in 1906, oxytocin has revealed itself to be the quintessential mammalian peptide, being part of a cascade of factors involved in childbirth and parental care. Synthesized in the hypothalamus, it functions both as a hormone—binding to receptors on cells in organs in the body such as the uterus (to trigger contraction) and breasts (to initiate milk letdown)—and as a neuromodulator, binding to receptors on neurons in the brain.

Animal research over the last 40 years has established that oxytocin mediates social behaviors. When group-living mammals encounter familiar conspecifics, oxytocin is released in the brain and motivates approach behaviors. The peptide also plays a critical role in maternal attachment, pair bonding, and even the maintenance of cross-species “friendships” that are valuable for protection, food sharing, and learning through play.

While scientists are just beginning to use neurological measurements to predict what people will do, the opportunity to improve people’s lives holds great promise.

Oxytocin is typically measured in cerebrospinal fluid in animals. When I hypothesized that oxytocin might explain human social behaviors, I did not want to subject people to spinal taps. An odd property of oxytocin is its simultaneous release in the brain from cells of the hypothalamus and in peripheral blood from the posterior pituitary. Taking advantage of this property, my lab developed a protocol to measure the change of oxytocin in blood that would reliably reflect the change in the brain. But oxytocin has a 3–5 minute half-life and relatively fragile chemical bonds, so we would have to draw blood rapidly, and keep it cold while it was processed. It was a tricky procedure to sort out, and for a month my arms were riddled with prick marks as I let my PhD student Bill Matzner, also a practicing physician, practice on me, the only test subject I knew would consent to multiple needle sticks. 

Once we learned how to handle the blood samples, we still needed a way to stimulate oxytocin release. Childbirth and breastfeeding would not do, but the animal literature had demonstrated that oxytocin release could be triggered by a social stimulus. So I turned to a senior colleague for advice. 

In the 1990s, members of Vernon Smith’s lab at the University of Arizona experimented with sequential money-sharing tasks to document cooperation between strangers even when there is an incentive to cheat. The task they developed separates participants and masks their identities, allowing them to act freely without fear of judgment as they interact with one another by computer. Each participant earns $10 for agreeing to join the experiment and is randomly paired off with another individual into dyads of decision maker 1 (DM1) and decision maker 2 (DM2). After comprehensive instructions, the computer prompts DM1 to send between $0 and $10 to DM2. Both DMs know that whatever DM1 chooses to send is removed from her or his account, and that three times the amount is added to DM2’s account. DM2 then gets a prompt showing the tripled amount received along with the new total in his or her account, and is prompted to transfer an amount between zero and the account total back to DM1. The chosen sum is removed from DM2’s account and that same value is deposited in DM1’s account. Participants never meet, and are paid their account balances in cash privately when the experiment concludes.

Both participants can walk away with more than their $10 show-up earnings if DM1 believes that DM2 will cooperate and if DM2 chooses to do so. Across many studies using this task, from Smith’s lab and other groups, that’s often what happens: most DM1s send some money to DM2s, and most DM2s return enough to bring DM1’s account balance to more than the $10 she started with. The consensus view by experimental economists is that the DM1 to DM2 transfer is a signal of trust. The DM1 transfer seems to say, using the southern California vernacular where we ran our studies, “Hey dude, we can soak these scientists for a bunch of cash. I trust that you understand why I’m sacrificing to grow the pie and will flip some dough back to me.” The back-transfer from DM2 to DM1 is then a measure of trustworthiness.

Smith, who later won the Nobel Prize in Economic Sciences for introducing experimentation to the field, told me that he had no idea why people cooperated with strangers on this task considering that DM2s can keep all the money without suffering any repercussions. I suspected that our human social nature, perhaps due to oxytocin, was the explanation. I hypothesized that DM1’s transfer of money, denoting trust in DM2, would cause an increase in oxytocin in DM2 that would spur him to reciprocate. The key to testing this hypothesis was measuring the change in oxytocin over the course of the experiment. 

In studies published in 2004 and 2005, we reported that 90 percent of DM1s sent some of their money to DM2s, and that the more money DM2s received, the larger their spike in oxytocin. The magnitude of the oxytocin rise was also highly correlated with the amount of money DM2 returned to DM1. It wasn’t the money itself that caused the oxytocin surge; a control condition, where the DM1-to-DM2 transfer was determined randomly, and the participants knew DM1 had no choice in the matter, produced no oxytocin response and few back-transfers to DM1s. Rather, the data indicated that it was the intention of trust from DM1—reflected in the choice of transferring money—that triggered the rise in oxytocin in DM2. We also measured nine other neurochemicals known to affect the release of oxytocin or the binding of oxytocin to its receptor; none of these correlated with oxytocin release or money transfers. It appeared that oxytocin alone was responsible for reciprocity.  

Putting oxytocin to the test

Scientists by personality and training are skeptics. Even before publishing our findings relating the endogenous release of oxytocin to trustworthiness, I was concerned about causation. To demonstrate that oxytocin caused trust-related behaviors, I had to get oxytocin safely into human brains. Synthetic oxytocin (trade name Pitocin) has a very safe drug profile; intravenous Pitocin is regularly used to speed up labor. But the evidence indicated that little or no intravenous oxytocin enters the brain. An intranasal oxytocin infuser, used to initiate milk flow for breastfeeding, was available in Europe but had gone off the market in the US, despite also being very safe. I found a Swiss pharmacy that would mail oxytocin infusers to me with a prescription, but after a year of corresponding with the US Food and Drug Administration (FDA), I was told that I would not be allowed to import the drug. 

In 2003, Swiss graduate student Markus Heinrichs, now a professor at the University of Freiburg, sent me his doctoral dissertation that used intranasal oxytocin administration to study social stress in humans. Heinrichs had shown that oxytocin reduced stress responses when participants interacted with one another prior to a public speaking task. Desperate to show oxytocin was causally related to trust, I called Markus and suggested a collaboration: he had the intranasal oxytocin infusers and I had the trust task. We put a group together and ran the study in Switzerland. 

Think of synthetic oxytocin as inducing the same physiological effect one would experience when meeting a friend. While in our previous experiments, DM1s did not get a social signal prior to making a choice and thus did not show a change in oxytocin, we expected that an oxytocin boost in the form of intranasal infusion would influence DM1s’ behavior. Sure enough, our experiment showed that oxytocin more than doubled the number of DM1s who sent all their money. Overall, DM1s who received an infusion transferred an average of 17 percent more money to the DM2s in their dyads than those given a placebo. At the same time, the additional oxytocin had no effect on DM2 behavior. This was not entirely unexpected, because there is already a strong desire to reciprocate trust, as evidenced in our previous experiments; more oxytocin, it seems, could not push return transfers even higher. Our team published these results in 2005 in Nature.

Shortly thereafter, I obtained approval to use intranasal oxytocin in the US. Our studies of endogenous oxytocin and exogenous oxytocin infusion showed that the molecule affects many prosocial behaviors such as generosity toward strangers and charitable donations. These findings were extended by other labs showing that oxytocin administration increased empathy, did not make people gullible, and had context-dependent effects. While patients with social anxiety, depression, and autism have dysregulated oxytocin, administering oxytocin in these patients does not seem to help alleviate the symptoms, likely because it does not change the underlying neural circuitry. My team also showed that endogenous and exogenous testosterone, known to inhibit oxytocin release, reduced trustworthiness and generosity in money-sharing tasks. Oxytocin appeared to be a critical part of the fabric of human sociality

But we wanted to see how far we could go. After my experience watching Million Dollar Baby, I wondered if more than just personal interactions could trigger the release of oxytocin. 

Capturing Immersion

Immersion is a neurological state of attention and emotional resonance that predicts what people will do after an experience, often with 80 percent or greater accuracy. We identified it by comparing neural activity in people who took an action after an experience versus those who did not.

© MICHELLE KONDRICH

Participants viewed a video about “Big” Ben Bowen, who suffered from an aggressive brain tumor at age two and was featured in a fundraising campaign from St. Jude Children’s Research Hospital. As participants watched Bowen’s story, we measured attention and emotional responses using brain activity as measured by electroencephalography (EEG) as well as multiple signals from the peripheral nervous system. We also drew blood from participants before and after they viewed the video to measure changes in oxytocin, cortisol, and adrenocorticotropic hormone (ACTH). One-half of the hundreds of people who viewed the video donated money to St. Jude. Our analysis predicted who would donate with 82 percent accuracy.

See full infographic: WEB | PDF

Immersion in an experience

Entertainment only works if one cares about the characters in the narrative. Movies put this effect into overdrive by combining visuals, music, and emotional displays. Neurologically, it is odd that people who are cognitively intact and sitting in a movie theater (or on an airplane) will cry or laugh at a flickering image. I wanted to know if the movies and videos that elicited such emotional responses were provoking oxytocin release.  

In order to rule out the random neurochemical changes that are part of the brain’s housekeeping, my lab and I designed experiments that included a behavioral task. If the neurochemical changes we observed were a response to the video and not background noise, they should correlate with choices made by participants after viewing the videos. 

Our first study used a fundraising video we obtained with permission from St. Jude Children’s Research Hospital. We cut out two 100-second segments of a father with his two-year-old son who was dying of brain cancer. The narrative version of the video showed the father talking about how it felt to know that his son was dying. The control video showed the same father and son at the zoo. We paid people to participate and obtained blood samples before and after they viewed the video. After the second blood draw, participants had the option to donate some of their earnings to St. Jude. This was done in private so people would not feel pressured to give. The narrative version of the video caused a spike in oxytocin while the control did not. Furthermore, the increase in oxytocin correlated with the amount of empathic concern participants reported for the father and son. But oxytocin alone was not sufficient to predict whether participants would donate.

Going into the study, we had decided to measure neurochemicals associated with arousal, knowing that very high arousal can inhibit oxytocin release. It turned out that arousal was a critical factor in determining who gave money to St. Jude. The analysis showed that, in addition to an increase in oxytocin, participants who donated showed positive arousal—initially, a rise in cortisol, and later, an increase in the faster-acting adrenocorticotropic hormone (ACTH). The data hit us with a new insight: it seemed that participants had to both pay attention to the video—i.e., be aroused—and be emotionally engaged with it to donate. 

In order to establish whether or not oxytocin plays a causal role in people’s decisions to donate to charity, our next study administered synthetic oxytocin or placebo to people before they watched 16 European public service announcements (PSAs) in English. The videos warned viewers about public health issues such as heart disease and drunk driving. Many were funny, some were sad, and others simply stated facts. Participants earned $5 after watching each video and could donate to a US-based charity that worked on the highlighted problem. Sure enough, participants who received oxytocin donated 56 percent more money to 57 percent more charities and reported more concern for the people shown in the PSAs compared with placebo recipients. We recruited a new batch of people and had them watch one of the 16 PSAs and took blood samples before and after. As in our previous study, donations were made by those who had increases in oxytocin and ACTH. We coined the term “immersion” to denote the neurological state of attention and emotional resonance during an experience that results in an observable behavior. 

To more fully understand what stories such as the St. Jude video do neurologically, we needed higher-frequency data. Blood draws before and after viewing cannot capture second-by-second responses in the nervous system. In 2011, the US Defense Advanced Research Projects Agency (DARPA) had launched a program called Narrative Networks that supported neuroscience research into persuasive communications. I had met the program officer, William Casebeer, at several conferences and he invited me to submit a proposal for funding. 

Once funding was secured, we returned to the St. Jude video and measured myriad physiological signals including cardiac rhythms, vagus nerve activity, and electrical conductance of skin to capture arousal and emotional responses without blood draws. The data showed that the attentional response occurred first while the emotional response followed typically 10–15 seconds later. Combining arousal and the effect of oxytocin into a measure of immersion, the pattern looked like a classic narrative arc, with the intensity of immersion peaking at the video’s climax and declining as the video resolved. Our subsequent studies of hundreds of audio and video stories showed that the narrative arc is an effective way to sustain immersion and motivate actions—which makes sense, as it’s been used to teach and entertain for thousands of years.

Think of synthetic oxytocin as inducing the same physiological
effect one would experience when meeting a friend.
 

Our contract with DARPA required that we predict with at least 70 percent accuracy who would donate after the video using only neurological data. Building statistical models from the electrical signals that constitute immersion, our predictive accuracy in 2015 was 82 percent. Modeling second-by-second data, we found that those who donated to St. Jude had a more pronounced spike in immersion at the peak of the narrative arc compared with non-donors. Measuring immersion is a way to understand, and predict, why people do what they do. 

Predicting people

In 2013, my Google alert for the word “oxytocin” picked up a video from Cannes, France. In it, Josy Paul, chairman of the India division of the global advertising agency BBDO, said that BBDO’s ads were so creative that “they caused the brain to make oxytocin.” I was intrigued. And skeptical. 

Had BBDO really measured oxytocin using blood draws? A few searches led me to Paul’s email, and I sent a query. They were “guessing” about oxytocin, he said, but they were interested in doing a test. I explained that immersion, not just oxytocin, was the best predictor of actions that I had found, and that I could measure it with wireless sensors. BBDO executives cooked up a plan to hold my feet to the fire: a blinded prediction. 

Here is how it went. BBDO sent me 18 TV commercials they had created for six different brands. There were three commercials each for Snickers candy bars, Cesar dog food, AT&T phone services, Visa credit cards, and two beers, Guinness and Bud Light. BBDO’s clients had already ranked the commercials by the sales bumps they had produced—information that BBDO withheld from us while we analyzed people’s immersion levels as they watched each commercial. 

A blinded prediction is the ultimate test of accuracy because there is no way to cherry-pick the data or do esoteric statistical analyses to improve the forecast: either immersion predicted sales or it did not. We recruited 61 participants to watch the BBDO commercials. Immersion correctly identified the ads that produced the largest sales bump for five of the six brands. Moreover, we found a statistically significant linear relationship between immersion and sales bumps: as immersion increases, so do sales. BBDO was thrilled, as was I.

The BBDO study gave me the confidence to scale up data collection. I started a company called Immersion Neuroscience and, together with my collaborators, built a software platform that allowed anyone to measure immersion. The Immersion Neuroscience software lives in cloud servers and takes the signal from popular wearables that measure cardiac activity, using algorithms we wrote to infer neural states in real time. We launched the platform in 2018, and since then, clients have used it to measure immersion in ways we had never thought of. 

Rather than lab experiments, Immersion’s clients are performing field studies. These studies have shown that immersion can identify top-rated reality TV shows with 88 percent accuracy and that immersion while listening to music three months prior to release had a nearly perfect correlation with post-release Spotify streams. Client usage has also found that information recall two weeks after a presentation has a high positive correlation with immersion. Since we launched our platform, a major professional services company, Accenture, has been measuring immersion during the training they provide to employees and has used immersion data to ensure all learners benefit from training. Real-time feedback on immersion is increasingly important as training and education go virtual and instructors are not in the same room with learners. 

I have had the privilege to take the basic science we have done on the behavioral effects of oxytocin and create a tool to predict what people are likely to do. Immersion seems to capture the value of experiences. When experiences are valued, they are remembered, acted on, and shared with others. Oxytocin is only part of the neurological process that leads to actions, but the research from my lab and many others during the last 20 years shows it is an important signal of the value that people derive from experiences with social content. 

Paul J. Zak is a professor and founding director of the Center for Neuroeconomics Studies at Claremont Graduate University in California. He is also the founder and Chief Immersion Officer of Immersion Neuroscience, which launched the first “Neuroscience as a Service” platform that enables clients to predict human behavior by measuring immersion.