Understanding Human Decision-Making: Neuroeconomics

In 1670, Blaise Pascal took what many see as the first steps towards an analytical understanding of human decision-making. He argued that when we make choices we should weigh the value of alternatives quite literally–assessing the worth of each of the options before us and comparing them directly. His work laid the foundation for the social scientific study of how we, as humans, decide. For the three hundred years or so that followed the birth of scholarship on decision-making, however, biologists largely avoided the subject. The physiologists of the late 20^th century focused their attention on sensory and motor systems, brain systems coupled to the external environment more directly than the cognitive systems associated with decision-making.

That changed in the first years of the 21^st century during which the study of human and animal decision-making by neuroscientists has blossomed. In 1990-91, just 6 academic papers used the word decision in their title and brain in their abstract. In 2016-17, that number had jumped to 213. It seems today that the study of human and animal decision-making is everywhere, so now seems a good time to take stock of both what we have learned and what we might expect to learn in the years to come.

Taking It from the Top…

Taking their cue from Pascal, economists, who focus on the study of human decision-making, have long hypothesized that when we make decisions we find some way to represent, in a numerical sense, what a given option is worth to us. In the language of modern scholarship, these are the “subjective values” that guide each of our individual and idiosyncratic decisions. The criticality of the idea that we can assign single numerical values, reducing many dimensions to one, cannot be overstated. Despite the popular saying, when we go to the supermarket we really do compare apples and oranges, and often buy one or the other depending on what we personally value. It is this ability to choose between disparate options is what drives economists to the notion that we can place a single relative value on any “good,” for the purposes of comparison. Following this line of argument, economists have argued for centuries that we behave exactly as if we placed a subjective numerical value on goods like apples and oranges when making comparative choices. Of course they acknowledged that our hunger levels and even our age, amongst other things, influence those subjective values. But the core idea, that somewhere in our brains we represent subjective value as a numerical quantity for comparing options, has been a persistent feature of social science for centuries. More recently, behavioral economists have broadened our understanding of these other things that influence valuation, including numerical representations of one’s expectation, one’s fear of losses, and subjective notions of probability, into their models–while retaining this notion of a subjective value.

Hunting for Subjective Value in the Brain

This long-standing tradition in economics led neuroscientists in the early 2000s to begin a concerted search for the neural representation of subjective value, a search for the brain loci where we assess our alternatives in the way originally proposed by Pascal. My lab produced some of the first evidence for such a subject valuation signal in the posterior parietal cortices of monkeys (Platt et al, 1999) and our work was quickly followed by closely related studies in humans that identified value-related signals in the human frontal cortex (Delgado et al., 2000; Knutson et al., 2000; Elliot et al, 2000). Evidence then accumulated rapidly that three or four areas in the human brain (the ventral striatum, the ventromedial prefrontal cortex, the posterior cingulate cortex, and the posterior parietal cortex) contained robust value-related signals appropriate for guiding decision-making.

Perhaps surprisingly, the suggestion in that early work that it was subjective value which was represented was highly controversial. At the time, many scientists argued that these value representations were much more likely to represent the true external value of the rewards offered to humans or monkeys in experiments, rather than representing the idiosyncratic subjective values that economists had hypothesized guide people’s choices.

In 2007, my colleague Joe Kable and I put much of that controversy to bed by studying how humans idiosyncratically value delayed monetary rewards. We found, unsurprisingly, that some humans are impulsive, valuing rewards only in the moment, while others are patient, valuing rewards long into the future. More importantly, we found that these highly subjective behavioral valuations could be predicted from brain activity levels in two critical brain valuation areas: the ventromedial prefrontal cortex and the ventral striatum.

If you were idiosyncratically impulsive in your behavior towards a particular reward, your ventral striatum was just exactly as idiosyncratically impulsive in its numerical valuation of that reward. In a similar way, Kenway Louie and I (2010) showed that the exact idiosyncratic numerical valuation read out from the parietal cortex of monkeys precisely predicted the subjective choices that monkey would make. Subsequent studies reinforced this conclusion in several ways, and it is now pretty much dogma that at least some areas of the brain encode the subjective values that guide our choices.

We are now even beginning to see evidence that the structure of our brains, not just the activity in the brain, influences how we generate subjective values. To see how this works consider how differently each of us might view a lottery ticket that offered a 50 percent chance of winning $100. Some might be willing to trade that ticket for a sure win of $40 despite the fact that the average value of such a lottery ticket was $50 (0.5 x $100). Economists refer to that as risk aversion, and it turns out that the thickness of a tiny patch of parietal cortex does quite a good job predicting how risk averse each of us is (Gilaie-Dotan, S. et al., 2014). Indeed, it seems that changes in risk aversion as we age (we famously become more risk averse as we age) correlated better with the age-related thinning of this brain area than with age itself (Grubb et al., 2016).

Perhaps the most exciting finding in this general domain is more causal, the demonstration that changing the activities in some of these areas actually changes what we choose. Daria Knoch, Ernst Fehr, and their colleagues (2006) famously showed that how highly an individual values fairness to others in monetary settings can be systematically manipulated by activating or deactivating brain areas that code subjective value.

Behavioral Economics and the Brain

At the same time that neuroscientists have made such huge strides towards understanding where in the brain subjective value is represented and some smaller strides towards understanding how it is constructed, many social scientists have been moving in a different direction. Chastened by the limitations of economic models from the mid-twentieth century, these psychologists and economists have begun to develop new models that rely on additional variables that play a role in setting subjective value. This has given rise to a whole new subdomain in economics and psychology called behavioral economics.

Behavioral economics was popularized, in large measure, by Stephen Dubner and Steven Levitt’s massive bestseller: Freakanomics. But behavioral economics is much more than a series of anecdotes. It is an effort to identify the critical elements from which the momentary subjective values that guide choice are built. Nearly all of these elements are efforts to capture how context influences what we value. Offering someone a raise of $5000 can be great if they expect a lower raise, but can cause them to quit if they expect more.

Today this kind of context dependency is captured by the idea that all of our choices are made relative to a reference point (Koszegi and Rabin, 2006). Neurobiologically, there is now growing evidence that we can see in the brain the reference point and its interaction with subjective value (Tom et al., 2007). Even more, there are developing theories that use core ideas from neuroscience to explain the sources of both reference points and context dependence in neurobiological representations (Glimcher, 2015).

The Path Forward

Many scholars working in the field today are quite optimistic that the neuroscience and economics of decision-making are now really beginning to converge. Benedetto DiMartino, Colin Camerer, and their colleagues (2013) recently demonstrated that brain scanners can be used to predict behavior during stock market bubbles. Models rooted deeply in the history of neuroscience are beginning to appear in economics journals. While it is safe to say that much work remains to be done, it also is clear how much has been accomplished.

In 1992 at the Society for Neuroscience’s annual meeting, there were just two presentations, out of more than ten thousand, that dealt with the neuroscience of decision-making. At the 2016 meeting “decision-making” was one the 10 curated itineraries provided by the Society to guide visitors through the most popular subjects presented at the meeting. The future seems very bright for those of us neuroscientists who hope to truly understand how we do what Pascal argued we must do.

Further Reading:

Glimcher, PW and Fehr, E. (2014) Neuroeconomics: Decision Making and the Brain. New York: Academic Press.