Three Correlated Essays on the Neural Foundations of Economic Decision-Making

Citation

Hsu, Ming (2007) Three Correlated Essays on the Neural Foundations of Economic Decision-Making. Dissertation (Ph.D.), California Institute of Technology. doi:10.7907/JE01-F098. https://resolver.caltech.edu/CaltechETD:etd-02022007-163425

Abstract

Paradoxes are useful in science because they hint at errors or inconsistencies in our models of the world. In this thesis, I study two well-known and long-standing paradoxes in decision theory from the point of view of neuroeconomics. This approach combines tools and ideas from economics and neuroscience, and tries to understand the neural mechanisms and the causal structures behind these paradoxes.

Since its introduction in Ellsberg (1961), the Ellsberg Paradox has been one of the most studied violations of subjective expected utility theory (SEU). In one version of the paradox, a decision-maker is confronted containing two urns with 100 balls that are either red or blue. In the first (risky) urn, she is told there are 50 red and 50 blue; whereas no further information is given about the second (ambiguous) urn. A commonly observed choice pattern is for decision makers to choose to bet on both red and blue in the first urn. Clearly, if probabilities are additive, such rankings are inconsistent with SEU.

First, I present brain imaging that shows that the brain treats risky and ambiguous choices differently. This is done through the use of functional magnetic resonance imaging (fMRI), a method that measures brain activity indirectly through blood flow. I find evidence that the brain regions respond differently to ambiguity and risk. Furthermore, the region that is correlated with expected money value of choices are activated more under risk than ambiguity, confirming that expected utility of ambiguous gambles are lower than those of equivalent risk gambles. Finally, the temporal relationship between the regions suggests a network where one brain region signals the level of uncertainty (amygdala), sent through another region (orbitofrontal cortex), and increases (decreases) expected utility of the choices, represented in the activity of a third region (striatum).

Brain imaging results, however, is limited by its correlational nature. To assess necessity, if a particular brain region causes a certain behavior, taking it out should remove that behavior. Conversely, to assess sufficiency, stimulating the brain region should create that behavior.

In the former, I study patients who have damage to the orbitofrontal cortex (same region found in the brain scans). I find that these patients were both ambiguity- and risk-neutral. This compares to ambiguity- and risk-averse behavior of patients with damage to other parts of the brain not found in the brain scans, similar to normal individuals. This confirms the idea that specific brain regions are necessary for distinguishing between risk and ambiguity. In the latter, I activate amygdala of (normal) subjects through mild electrical stimulation (a method known to elicit activation of the region). This allows us to test whether this method increases the ambiguity/risk aversion of subjects.

The third chapter studies the Allais Paradox and the probability weighting function. The fact that people do not appear to weight probabilities linearly as dictated by subjective expected utility theory has been known since the 1950s. More specifically, people have been found to overweight small probabilities, and underweight large probabilities. This chapter has two goals. First, I attempt to find the neural correlate of the probability weighting function: that is, is the probability weighting function as discussed in the decision theory literature found in the brain? Second, I posit a hypothesis for the generation of the probability weighting function with data from psychophysics and neuroscience. Together they shed light on how the brain encodes probabilities as a physical quantity as well as how it might combine decision weights and rewards to calculate expected utility.

Thesis Files