Pricing Information and Data

Abstract

As the Internet makes a huge amount of information available, our lives have been greatly transformed by information. We use all kinds of information to make decisions: we use directions provided by Google Maps, ratings provided by Yelp, etc. However, useful information does not come for free. It takes human labor to collect and aggregate the information. Despite the fact that we live in the age of information explosion, we are still in need of good information/data. For example, one of the major challenges in machine learning is the lack of high-quality data.

Recent years have witnessed a growing information economy. Different data are collected, aggregated, and sold to their users. This information economy faces a fundamental question: how should we price the information or data that are being collected/distributed? The pricing of information/data is challenging for several reasons: (1) unlike physical goods, information/data can be easily fabricated or modified; (2) it is not clear how to decide the value of a piece of information/a data set; (3) information can be partially revealed. This thesis explores the pricing of information in the face of these challenges by focusing on two important parts of the information economy: the acquisition of information, and the sale of information. We zoom in on three scenarios that are frequently encountered in practice.

The first part focuses on pricing the information collected from self-interested agents when the information cannot be verified. Human-generated information is increasingly prominent and important. Information provided by human workers usually cannot be verified, since the objective truth is hard to access or even does not exist. The pricing problem is thus challenging. We explore the design of pricing mechanisms that collect honest information from self-interested agents.

The second part investigates a monopolist information seller's pricing strategy to maximize her expected revenue. One key difference between information products and physical products is that information can be revealed partially. Because of this property, the information seller has much more flexibility when designing the selling mechanisms. We investigate different types of mechanisms that an information seller can use to sell information products.

In the last part, we aim for data pricing methods that give cost-efficient outcomes for a data collector. When the data do not come for free, the data collector may want to wisely allocate her budget to get the most useful data. The usefulness of data depends on the specific problem in which the data are used. We design cost-efficient pricing mechanisms for two important classes of problems: linear optimization and statistical estimation.