Belief patterns with information processing

Federico Vaccari

2024-11-27

Abstract:This paper presents a model of costly information acquisition where decision-makers can choose whether to elaborate information superficially or precisely. The former action is costless, while the latter entails a processing cost. Within this framework, decision-makers' beliefs may polarize even after they have access to the same evidence. From the perspective of a Bayesian observer who neglects information processing constraints, the decision-makers' optimal behavior and belief updating may appear consistent with biases such as disconfirmation, underreaction to information, and confirmation bias. However, these phenomena emerge naturally within the model and are fully compatible with standard Bayesian inference and rational decision-making when accounting for the costs of information acquisition.

General Economics

What problem does this paper attempt to address?

The problem that this paper attempts to solve is: In the process of information acquisition, how can decision - makers still have the phenomenon of belief polarization when facing the same evidence? Specifically, the paper explores how decision - makers can make optimal information - acquisition decisions according to prior beliefs when they can choose to process information at different costs (i.e., shallow processing or precise processing), and analyzes the impact of these decisions on belief updating.

### Problem Background

1. **Cost of Information Acquisition**: Decision - makers can process information shallowly for free or choose to pay a certain cost for more precise processing.
2. **Differences in Prior Beliefs**: Different decision - makers may have different prior beliefs, which will affect their choices of information processing.
3. **Belief Polarization**: Even when facing the same evidence, decision - makers with different prior beliefs may make different information - processing decisions, thus leading to further divergence of their beliefs.

### Research Objectives

The goal of the paper is to explain why decision - makers still have the belief polarization phenomenon after obtaining the same information through a fully rational Bayesian framework. The author hopes to reveal the following points through this model:

- How decision - makers' prior beliefs affect their choices of information processing.
- How the cost of information processing affects decision - makers' belief updating.
- Under the consideration of the cost of information processing, how standard Bayesian reasoning and rational decision - making can naturally explain some common cognitive biases (such as confirmation bias, under - reaction to information, etc.).

### Main Contributions

1. **Established an Information - Acquisition Model**: This model allows decision - makers to choose whether to conduct more precise information processing at a higher cost.
2. **Analyzed Belief Patterns**: Studied the belief patterns of different decision - makers after optimal information acquisition, especially the conditions and probabilities of belief polarization.
3. **Explained Cognitive Biases**: Explained some common cognitive biases, such as confirmation bias and under - reaction to information, from the perspective of rational decision - making.

### Conclusions

The main conclusion of the paper is that belief polarization is not caused by irrational behavior, but because decision - makers with different prior beliefs make different information - processing decisions when facing the same evidence. Ignoring these information - processing incentives may make observers mistakenly think that decision - makers' behaviors show cognitive biases, while in fact these behaviors are rational and optimal when considering cost constraints.

### Formula Summary

The formulas involved in the paper are mainly used to describe decision - makers' willingness - to - pay function \( c_{\sigma_1}(p) \), which represents the maximum cost that decision - makers are willing to pay after observing signal \( \sigma_1 \). The specific formulas are as follows:

\[ c_{\alpha}(p) = \begin{cases}
0 & \text{if } p \in \left[0, \frac{1-\theta_1\theta_2}{1-\theta_1-\theta_2+2\theta_1\theta_2}\right], \\
\Delta U \cdot \frac{\theta_1\theta_2 p - (1-\theta_1)(1-\theta_2)(1-p)}{\theta_1 p + (1-\theta_1)(1-p)} & \text{if } p \in \left[\frac{1-\theta_1\theta_2}{1-\theta_1-\theta_2+2\theta_1\theta_2}, 1-\theta_1\right], \\
\Delta U \cdot \frac{\theta_2(1-p) - \theta_1 p - \theta_1\theta_2(1-2p)}{\theta_1 p + (1-\theta_1)(1-p)} & \text{if } p \in \left[1-\theta_1, \frac{\theta_2(1-\theta_1)}{\theta_1+\theta_2-2\theta_1\theta_2}\right], \\
0 & \te