Neuroforecasting¶
Annella Charee Tucker & Julia Lebovitz
Neuroforecasting research measures individual brain activity while completing decision making tasks to forecast aggregate responses. For our chapter, we plan to share a history of neuroforecasting, identify key brain regions and circuits, and comment on ethical concerns.
For this chapter, we interviewed Professor Brian Knutson, one of the pioneers and leading researchers in the Neuroforecasting field to better understand the importance of his / the field’s research.
Introduction to Neuroforecasting¶
In general to forecast is to make a statement about the future. Neuroforecasting is seeking to make a statement about future choices by analyzing the neural processes of individuals. This has been primarily accomplished by looking at brain circuits using fMRI. Neural data is collected from an individual while making decisions, and this information is used to forecast how an unrelated person will respond in the same situation. Researchers are seeking to use individual neural data to forecast aggregate choices.
The process that determines neural activity and the neural process that describes decision making are both explored within neuroforecasting. Functional magnetic resonance imaging, fMRI, is a technology that detects blood flow. Increased blood flow to a region of the brain indicates neural activity. Brian Knutson, the co-lead of the NeuroChoice Initiative of the Stanford Wu Tsai Neurosciences Institute has contributed to research using fMRI to scan the neural activity of study participants watching online videos. Knutson conducted neuroforecasting research intending to predict the potential popularity of online videos. Study participants simultaneously watched videos and were prompted by researchers with questions about the video. Questions were formulated to expose the participants’ opinions of the video. These verbal responses did not forecast the future popularity of videos, but brain responses did (Kubota). “If we examine our subjects’ choices to watch the video or even their reported responses to the videos, they don’t tell us about the general response online. Only brain activity seems to forecast a video’s popularity on the internet,” explained Knutson. The key takeaway? Brain activity matters and can reveal more than reported responses.
Brain activity reveals responses people are sometimes unable or unwilling to share. Verbal responses are not always accurate or insightful. The use of fMRI to gather data makes neuroforecasting applicable to fields such as marketing and economics. The concept actually evolved from neuroeconomics. Brain activity can predict the extent to which someone likes a product or what option someone might choose. A leading researcher in the field, Alex Genevsky predicted that “…we are getting close to a time when biologically based and neural measurements will play a significant role in mainstream marketing.”
Neuroforecasting research suggests that consumer choice may be more affectively-driven than reflective. It is important to note limitations with neuroforecasting research, primarily the research cited using fMRI technology. One major limitation comes with the expense of fMRI technology. Due to the cost and long amounts of time needed in the scanner, neuroimaging research has relatively small sample sizes. Further, as the machine is immobile, neuroforecasting research may not be able to actually replicate real-world scenarios.
Brain Regions and Circuits Associated with Consumer Choice¶
In this section, we will detail findings from individual neuroforecasting studies as well as a systematic review to demonstrate neural regions, and perhaps brain circuits, implicated in this field.
Fig 1: Brief Overview of Brain Areas Implicated in Neuroforecasting Research. More opaque regions (OFC and Insula) represent brain regions that are deeper in the brain. Generated Brain Image is a rough estimate to show a schema of literature findings. neurosynth.org used to generate model neural activity.
A landmark study in this field was conducted by Berns and Moore (2012) who used fMRI to measure neural activity when participants were exposed to different songs with the intent to track popularity. Two years after the researchers measured brain activity while listening to songs, they used brain activity and “liking” ratings to predict which songs became a “hit.” They found that while subjective ratings did not predict which songs may become hits, nucleus accumbens (NAcc) activity did. The average song likeability had significant path coefficients in both the orbitofrontal cortex (OFC) and NAcc; while the NAcc and OFC path was significant, only the NAcc predicted ‘liking.’ The medial prefrontal cortex (mPFC) + NAcc activity forecasted aggregate downloaded songs 2 years later. Their research was one of the first to demonstrate that neural activity may provide hidden information and could be more reliable than self-report measures.
Knutson and colleagues (2007) used event-related fMRI to measure neural activity while individuals completed tasks related to gain and loss anticipation. They found different brain regions correlated with and predicted different aspects of individual consumer choice. They found that: The nucleus accumbens (NAcc) activity was correlated with product preference; insula activity occurred when participants were exposed to excessive prices; the mesial prefrontal cortex (mPFC) activity preceded the purchase decision. These areas are important leading up to the purchasing decision and activation across the circuit occurs prior to an individual consumer’s decision.
Falk, Berkman, and Lieberman (2012) showed participants different ads from campaigns geared at quitting smoking to measure the individual and population-level success of each advertising campaign. They found that the mPFC was most highly correlated with and predictive of both individual and aggregate behavior change to stop smoking. The mPFC has also been found to be correlated with a number of implicit preferences and judgements as well as considerations of personally relevant future goals.
Tusche, Bode, and Haynes (2010) measured neural activity and self-report liking in two experimental groups – high attention and low attention – while individuals were exposed to cars. Insula and mPFC activity predicted consumer choice under both conditions, suggesting that these areas are important in consumer choice even when an individual was not closely attending to objects.
Kühn, Strelow, and Gallinat (2016) measured neural activity while individuals viewed six different chocolate bars. NAcc activity was associated with product preference. High prices were associated with both increased insula activity and decreased mOFC activity. Insula activity predicted a decision to not buy the product while NAcc and mOFC activity predicted decisions to buy the product.
In a systematic review, Knutson and Genevsky found that in 8 studies forecasting aggregate choice with functional MRI before 2017, the NAcc, mPFC, or combination of the two regions acted as an aggregate outcome forecaster. Further, the NAcc, mPFC, and amygdala helped to forecast individual choice. This suggests that subcortical regions (particularly the NAcc) coupled with cortical regions (particularly the mPFC) are implicated in consumer choice. Their review indicated that not all regions used to predict individual choice could also be scaled to predict aggregate choice. They developed the affect-integration-motivation (AIM) framework, which suggests that neural signals are first effectively evaluated, then integrated, and finally translated into a motivated choice to avoid or act. They specifically expect that “neural signals from evolutionarily conserved affective circuits are cortically integrated with individual contextually relevant concerns and then relayed to motor preparatory circuits that can support motivated choice behavior” (112). As noted in the interview with Dr. Knutson, this circuit may be considered motivational as the NAcc and mPFC are both innervated by dopamine, a neurotransmitter important for wanting. Yet, more studies are needed to fully elucidate this circuit and understand when and why neuroforecsting works.
Real-world applications of the research¶
In our interview with Dr. Knutson, he recognized a general lag between this research and “real-world” applications. He commented that he (and people he was worked with) have acted as consultants for some marketing companies, but so far, to our knowledge this research has not directly had real-world applications. This is likely due to how new the field of neuroforecasting is. What this research does show is that certain brain regions may be able to predict both individual and aggregate choice. For example, Dr. Knutson noted that images of smiling individuals may boost NAcc activity. Neuroimaging may inform the economic and marketing fields as it could provide insight into hidden information and perhaps why people act irrationally at times.
The potential impact of neuroforecasting is interdisciplinary. Since the birth of the field in 2010 with the article “A Neural Predictor of Cultural Popularity” by Sarah Moore and Greg Berns studying neural imaging to predict the popularity of songs, the prospect of using neuroforecasting to sell has been at the forefront. A study using Kickstarter, a crowdfunding platform, investigated the success of project funding campaigns based on the neural responses of investors. Brain activity proved to be a better indicator of project funding success than reported responses of the study participants (Genevsky et al., 2017). Marketers are seeking to get into the minds of their consumers, and brain imaging allows them to understand decision-making processes.
Another field to be implicated by neuroforecasting is economics. Economists use models to explain patterns and trends, but these models are sometimes limited by the assumption that humans are ‘rational’ beings. Neuroforecasting models may help uncover hidden information that could help inform economic models.The intersection of neuroforecasting and economics supports economists in building models that take into account human behavior. While the intersection of the two disciplines is interesting, an economist does not necessarily have to become a neuroscience expert. “An economist who examines neural variables would not necessarily require extensive knowledge of neural processes; instead, he might simply rely on neuroeconomics to identify and collect the relevant data.” (Bernheim, 2013)
Ethics Of Neuroforecasting¶
While neuroforecasting has interesting applications, it also has potentially harmful implications. The insight brain imaging using fMRI can offer is tremendous, and in the wrong hands could be detrimental. One concern is how “sin” industries like alcohol and tobacco could utilize forecasting. Knowledge about addiction can be abused by the profit-driven industries and perpetuate addictions. Industry in general and the commercialization of brain imaging is a major issue as well. As marketing professionals discover how the information can be applied in practice they will likely be interested in developing in-house studies. Further, an individual’s privacy is of concern as neuroforecasting suggests ‘hidden information.’ However, using neuroforecasting to predict aggregate outcomes and not individuals has tremendous marketing / economic potential.
References¶
Berkman, E. T., & Falk, E. B. (2013). Beyond Brain Mapping: Using Neural Measures to Predict Real-World Outcomes. Current directions in psychological science, 22(1), 45–50. https://doi.org/10.1177/0963721412469394
Berns, G. S., & Moore, S. E. (2012). A neural predictor of cultural popularity. Journal of Consumer Psychology, 22(1), 154–160. https://doi.org/10.1016/j.jcps.2011.05.001
Bernheim, B.D. (2008). Neuroeconomics: a sober (but hopeful) appraisal. AEJ: Microeconomics.
Falk, E. B., Berkman, E. T., & Lieberman, M. D. (2012). From neural responses to population behavior: neural focus group predicts population-level media effects. Psychological science, 23(5), 439–445. https://doi.org/10.1177/0956797611434964
Genevsky, Alexander et al. “When Brain Beats Behavior: Neuroforecasting Crowdfunding Outcomes.” The Journal of neuroscience : the official journal of the Society for Neuroscience vol. 37,36 (2017): 8625-8634. doi:10.1523/JNEUROSCI.1633-16.2017
Knutson, B., & Genevsky, A. (2018). Neuroforecasting Aggregate Choice. Current directions in psychological science, 27(2), 110–115. https://doi.org/10.1177/0963721417737877
Knutson, B., Rick, S., Wimmer, G. E., Prelec, D., & Loewenstein, G. (2007). Neural predictors of purchases. Neuron, 53(1), 147–156. https://doi.org/10.1016/j.neuron.2006.11.010
Kühn, S., Strelow, E., & Gallinat, J. (2016). Multiple “buy buttons”in the brain: Forecasting chocolate sales at point of-sale based on functional brain activation using fMRI.NeuroImage, 136, 122–128. doi:10.1016/j.neuroimage.2016.05.021
Tusche, A., Bode, S., & Haynes, J. D. (2010). Neural responses to unattended products predict later consumer choices. The Journal of neuroscience : the official journal of the Society for Neuroscience, 30(23), 8024–8031. https://doi.org/10.1523/JNEUROSCI.0064-10.2010