Essays in information noise and information acquisition in communication games

Abstract

This thesis consists of three independent studies. The first study works on the effect of confirmatory bias. The second study analyses the effect of ambiguity in the communication. The third study discusses two-sender cheap talk model with endogenous information acquisition.

In the first study, it develops formulations to capture the effects of confirmatory bias in two-sender models. In these models, the information obtained by the two senders is conditionally independent and endogenously determined by their investment choices. Based on the investment levels they choose, the senders strategically direct messages to the receiver. This studies main purpose is to explore the effects of confirmatory bias in different models by comparing the differences in senders' investment incentives with and without the existence of confirmatory bias. Rather than having an exclusively negative influence on the players, as commonly believed, the results of this study suggest that it is also possible for confirmatory bias to be beneficial.

In the second study, the term \ambiguity" is narrowly defined as the phenomenon whereby a message is changed because of factors that cannot be controlled by the players. The results show that ambiguity has positive effects that can improve the truth-telling probability and utility of both players. These effects are observed when different cost types are applied. The positive effect of ambiguity generates a preference for ambiguity among the players. This strong effect on the improvement of truth-telling holds even when the sender's type is assumed to be uncertain. Overall, the results demonstrate the strong power of ambiguity in communication and that this power is beneficial for both players. This study provides theoretical support for making use of the inevitable ambiguity in communication, which can help people improve their payoffs.

In the third study, it examines a two-sender cheap talk model in a one-dimensional setting, in which two experts with either the same or opposite preferences choose their levels of investment in information, and communicate with an uninformed decision maker. The information obtained by the two experts is conditionally independent and affected by the investment levels chosen by the experts. After investing, the experts report favourable information and can cheat at some cost if unfavourable signals are observed. When the senders communicate with the receiver simultaneously, the original expert tends to increase his investment in information acquisition if the second non-strategic expert has the opposite bias, and decrease it if he is biased in the same direction. If the agents play a cheap talk game as described in this paper, the cheating decisions made by two oppositely biased senders are strategically complementary. Their investment levels are also strategically complementary as long as some strict conditions are satisfied. Conversely, when the two senders have the same preference, both their cheating decisions and their investment choices appear to be strategically complementary, with no further restrictions. The findings show that in the cheap talk model with endogenous information acquisition, the relationship between the two senders strategies is dependent on their preference relationship.