This dissertation studies communication and information design in strategic settings. In each chapter, I develop a theoretical framework to study an interaction where a decision-maker who lacks access to information relevant for her decision has to rely on one or more interested parties to provide it. My focus is on examining how certain naturally arising constraints faced by agents in these interactions matter for the degree of information transmission, and for welfare of agents.

In the first two chapters, I (with my co-author Mark Whitmeyer) revisit the classic strategic problem of two sellers that compete to sell to a consumer by manipulating the availability of information on product quality. In this framework, we examine the role of the consumer's limited ability to process information. Accordingly, we model the consumer as incurring a cost to process more information, and actively deciding how much of available information about each product to process. Chapter 1 studies a baseline environment where our main finding is that sellers are encouraged to disclose more information than they would to a consumer who could freely process all available information. Chapter 2 shows that this finding is robust to various modifications of the environment (such as allowing for ex-ante heterogeneity between sellers). Our analysis illuminates two forces that produce this result: First, substitutability between the two sources of information; second, optimality (for the consumer) of deliberately ignoring some information. As we show, another notable implication of this finding is that the consumer might choose the better product with a higher probability than when she is unconstrained in her ability to process information. Our theoretical framework applies to a variety of settings where information is "complex," and the leading application is that of pharmaceutical companies disclosing drug quality information to doctors.

The third chapter is motivated by the problem of a school that decides how liberal its grading policy on a test should be, which in turn determines how much information about students' ability is available to potential employers. The school is constrained by the fact that the test may only generate information about a student's academic ability, whereas employers additionally care about other "soft skills." In my setting, the school may also write a letter of recommendation for the student after privately observing her soft skills--but unlike a grading policy, there is no rule determining the content of the letter. The goal of the school is to maximize a combination of benefits from good placements and long term reputation concerns. My main finding is that the school designs a less informative grading policy when it has the option of additionally writing the letter, than when it does not. Intuitively, this happens because the letter is credible and influential only when there is sufficient uncertainty about the student's academic ability: If academic ability is revealed to (probably) be high, the letter is understood to be exaggerated; if it is revealed to (probably) be low, the employer's optimal decision is already determined. More generally, my analysis delivers insights about the strategic tensions at play in a model of multidimensional communication with partial commitment.