The Effect of Priors on Risk Taking: Theory and Evidence
Speaker: Dr. Lawrence J. Jin (California Institute of Technology)
|
Topic: |
The Effect of Priors on Risk Taking: Theory and Evidence
|
|
|
Time&Date: |
12:00pm-13:30pm, 2018/8/3 |
|
|
Venue: |
Room A619, Teaching A |
|
|
Speaker: |
Dr. Lawrence J. Jin (California Institute of Technology)
|
|
| Abstract: |
Computational neuroscientists have proposed that many perceptual biases are the outcome of a Bayesian learning process. Under this theory, an agent holds a prior belief about possible stimuli she may encounter in a given environment. Due to the constraints on neural processing, she is only able to observe a noisy estimate of the true stimulus governed by a likelihood function. Bayesian learning dictates that the agent optimally combines her prior belief with the likelihood function to form an estimate of the true stimulus value. Experimental research has provided support for this theory in several sensory domains.
In this paper, we develop a theory and provide experimental evidence to support the same optimal process of Bayesian learning in the domain of risky choice. For both the theory and our experimental design, we vary the true stimulus distribution and derive the resulting likelihood function under efficient coding (Wei and Stocker, 2015). We then use Bayesian learning to derive the model’s implications on decision making. Our theory makes two novel predictions. First, the sensitivity of the probability of risk taking with respect to choice payoffs decreases in the volatility of the stimulus distribution. Second, the subjective value function derived from our theory exhibits diminishing sensitivity, a key component of prospect theory. We find support of both predictions using experimental data.
|
