The Limits of Neuroscience in Business

Neuromarketing. Neuromanagement. Neurofinance. Attempts to inject brain science into the business world have been nothing short of comprehensive — and, quite possibly, oversold.

Proponents of neuroscience in commercial settings argue that it can provide key insights that help explain consumer and employee behavior — insights that can ultimately be used to develop more appealing products and services. While that may be true to a certain extent, managers considering such applications aren’t served by broad claims that gloss over important caveats concerning the practical application of brain science.

This article aims to help managers critically evaluate vendor offerings based in neuroscience. In particular, there are three very significant issues that business leaders must understand if they hope to make informed decisions regarding investments in such products.

Issue 1: Proxies

One high-profile example of neuromarketing involves researchers who used brain data from individuals watching movie trailers to predict ticket sales. This led to claims that brain imaging could be used to refine marketing practices and boost box-office revenues.

To understand why such claims are questionable, let’s explore the problem with proxies.

A proxy is any indirect measure used to predict the outcome of something that is otherwise difficult or impossible to measure directly. For instance, baseball scouts often use minor league on-base percentage as a proxy to predict how frequently players will get on base in the majors. This indirect measure allows teams to make educated guesses without spending millions of dollars calling up every encouraging prospect.

For a proxy to be considered meaningful within any given field, it must meet three specific criteria:

• Reliability: Because correlations inferred from small data sets often prove inaccurate or erroneous, good proxies are derived from very large pools of data.
• Validity: Because it’s theoretically possible for any single experiment to generate highly variant outcomes (like when the European physics laboratory CERN erroneously detected particles moving faster than the speed of light), good proxies have been independently replicated.
• Utility: Given the finite nature of resources and capital, good proxies deliver a comparatively favorable cost-benefit ratio.

In baseball, minor league on-base percentage is reliable because it has been derived from thousands of players over several decades; it’s valid because it has been replicated using player data from dozens of independent international leagues; and it’s useful because it costs little to measure yet correlates well (0.53) with major league on-base percentage.

There’s no reason to assume that brain data can’t be effectively used as a proxy; we already use it in medical settings regularly. However, the brain research typically conducted for business purposes rarely meets the criteria of a meaningful proxy.

For starters, the majority of business-related neuroscience experiments draw their conclusions from very small data sets, which raises issues of reliability. The aforementioned movie trailer experiment involved only 58 individuals watching 13 movie trailers. Needless to say, from a statistical standpoint, it is incredibly difficult to establish reliability from such a small pool of data.

Next, neuroscientific studies are rarely replicated, which raises concerns about validity. Despite significant hype, the movie trailer research remains a one-off experiment. To be fair, the lack of replication in science is a well-known problem; researchers have far greater incentives to favor discovery over validation. However, given that the movie trailer case is frequently cited as one of the most impressive in the neuromarketing literature, it’s surprising that there have been no independent replication attempts.

Finally, the significant investment of time and money required to conduct neuroscientific research raises concerns about utility. Rough estimates suggest that the movie trailer experiment required $50,000 worth of specialized equipment, involved over 58 hours of setup and breakdown, and could be run with only two participants at a time. In this very same study, researchers had each participant complete a questionnaire that carried a nominal cost, required 20 minutes to complete, and could be done by hundreds of participants simultaneously.

Given those costs, what were the benefits? The brain activity results were found to correlate with total box-office sales at 0.52, while the questionnaire was found to correlate at 0.52 — exactly the same. Clearly, the questionnaire delivered a vastly superior cost-benefit ratio compared with the brain measurements.

Again, there’s no reason to assume that brain data can’t be effectively used as a proxy. However, if these measures are to become meaningful within business contexts, we must demand deeper data, more replication, and reasonable cost-benefit ratios.

Issue 2: Emergence

A primary function of fields like neuroeconomics and neurofinance is to leverage brain data to explain why customers act in the manner they do.

For example, in another high-profile study, researchers conducted the famous Pepsi Challenge inside a brain scanner. Based on the results, it was argued that in the presence of popular branding, the dorsolateral prefrontal cortex (DLPFC) of the brain can override sensory processing, compelling consumers to select a soda on the basis of historical preference rather than actual taste.

The principle of emergence helps explain why this explanation is not entirely credible. Emergence tells us that when many simple entities interact within a shared environment, novel and more complex behaviors can arise that are neither present in nor predictable by the simple entities themselves. Emergence embodies the common adage “The whole is greater than the sum of its parts.”

In the context of neuroscience, emergence argues that mental activity derives from the interaction of the nervous system with all other bodily systems (such as digestive, cardiovascular, and respiratory) and accordingly cannot be explained or predicted by any one of these systems in isolation. In other words, states of the brain cannot be used to accurately measure states of the mind.

When neuroscience is oversold, it ignores this complexity and inverts logical arguments. For example, imagine we know that brain area X is active when a person makes an economic decision, and we know that Betsy is making an economic decision. From this we can conclude that Betsy’s brain area X is active.

Now, let’s invert this sequence. Imagine we know that brain area X is active when a person makes an economic decision, and we know that Betsy’s brain area X is active. From this, we cannot conclude that Betsy is making an economic decision.

The problem here is, due to emergence, every brain region is correlated with myriad behavioral outcomes. For example, the amygdala activates when a person is experiencing joy, fear, disgust, anger, sadness, memory formation, facial recognition, bodily coordination, or a dozen other mental processes. Simply knowing that a person’s amygdala is active tells us nothing about what that person is doing, thinking, or feeling.

In order to guess at the meaning of brain activity, we must collect behavioral data. For instance, to postulate that a person’s amygdala activation is related to joy, fear, or sadness, we must first ask that person what they are currently feeling. However, this leads to an awkward circularity: If behavioral data is required to make sense of brain data, and if behavioral explanation is the ultimate goal of a field like neuroeconomics, then what role does the brain data truly serve?

To see this circularity in action, let’s return to the Pepsi Challenge study cited above. Recall that the researchers argued the DLPFC can compel people to choose a soda on the basis of historical preference over pure taste.

According to the principle of emergence, there’s no way this explanation could have been derived from the brain data alone, because the DLPFC is linked to an array of cognitive functions, including memory updating, regulating emotions, lie telling, motor planning, syllogistic reasoning, and more. Accordingly, the brain data alone tells us nothing specific about consumer choice.

In order to guess at the meaning of this brain data, the researchers were inevitably forced to collect behavioral data by asking participants to verbally report which soda they preferred after drinking each. Importantly, it was this behavioral data that ultimately allowed the researchers to put forth the explanatory concept most relevant to businesses: that brand preference can override taste preference. This is why the original Pepsi Challenge conducted in 1975 arrived at the exact same explanation for consumer behavior without ever considering the brain: Due to emergence, neuroscience was able to contribute little more than a confusing acronym and an underdetermined mechanism to the ultimate explanation of why consumers act in the manner they do.

Issue 3: Translation

There are untold examples of people trying to use brain data to develop practically useful principles for managers. However, this process has often proved to be rife with pitfalls.

Take, for instance, a high-profile study involving oxytocin, a hormone linked to many cognitive processes, including stress modulation, social attention, and memory formation — in other words, emergence in action. The researchers used serological measures to demonstrate that increased levels of oxytocin correlated with increased trust among individuals playing a monetary exchange game. From that correlation, they derived the principle that managers should employ techniques that trigger oxytocin release among colleagues to bolster team cohesiveness.

To understand why this claim is hollow, let’s look at the problem of translation. Within science, translation simply refers to the process of using data from one field to derive practically useful principles within a different, unrelated field.

Here’s the rub: Effective translation requires researchers to collect data within the field they are attempting to translate into — a requisite that makes any borrowed data largely obsolete.

For example, when managers were urged to use oxytocin-boosting techniques to increase team trust, the majority were left asking, “How?” Nowhere in the original brain data is it clear what specific behaviors lead to the release of oxytocin, making it impossible to construct a practical principle based solely on this research.

In order to derive true utility for managers, it would have been necessary to collect behavioral data from the managers themselves — something these researchers did not do. However, let’s pretend the researchers did collect this data and found that team trust increased every time a manager referred to a colleague by name. From this, we could put forth a legitimate principle: Refer to your colleagues by name in order to trigger oxytocin release and boost trust.

Hopefully you noticed, however, that the original brain data becomes irrelevant to the development of this principle. In the end, knowing the specific hormone involved is of no real concern; utility is ultimately derived from the data collected in the very field we were trying to influence.

And herein lies the problem with translation: Knowing that trust is correlated with the release of oxytocin is of no value if the goal is understanding how to build trust. Accordingly, it’s safe to say that any practical principles derived from a combination of brain data and business data will rely exclusively on the latter.

Managing Expectations

The enthusiastic application of neuroscience across many diverse aspects of business is certainly impressive, but hopefully this article gives potential users pause and offers insight into what we can realistically expect from raw brain data.

For business leaders, it’s important to ensure that neuroscientific researchers clarify what measures they will be taking in advance and justify how their measures will specifically avoid the pitfalls of proxies, emergence, and translation outlined above.

For researchers, it’s important to publish academically before disseminating findings more broadly to facilitate effective peer review. Furthermore, researchers working within business settings should embrace the same careful language employed within academic settings, such as describing brain regions as “corollary to” rather than “causative of” an effect, in order to avoid misinterpretation.

Ultimately, if we are going to apply the tools of brain science to the domain of business, then we must also employ the same standards and practices that define the wider field of neuroscience. Proxies, emergence, and translation are issues that constrain all forms of brain research. Simply hitching the prefix “neuro-” to business functions like “finance” or “marketing” does not absolve anyone from needing to address these limitations.