Lori Andrews’s new book on privacy, I Know Who You Are and I Saw What You Did, has been getting a lot of press, and one of the most widely noted parts of her study has been the use of data aggregators for “behavioral credit scoring” by credit card companies and banks. Through the use of behavioral scoring, credit providers try to predict the creditworthiness of customers based on the payments of others with a similar behavioral profile. The outcry is based on a sentiment that this application of data aggregation is both particularly intrusive and that assessments of one’s creditworthiness are being improperly based on the activities of one’s peers. Both claims merit some thought, particularly considering what credit scoring tries to achieve. Credit, after all, is etymologically derived from trust. And assessments of creditworthiness are fundamentally about trustworthiness. (Side note: while financiers may be scratching their heads at the unwillingness of homeowners to walk away from bad mortgages and default, the moral dimension of debt and being a responsible borrower retains some force more generally.) Assessments of creditworthiness may be primarily based on financial metrics, but they implicate notions of character. The use of data aggregators to find new bases for credit scoring picks up where an old story left off.
Consider the major credit bureaus. Prior to the passage of the Fair Credit Reporting Act, Equifax (then known as the Retail Credit Company) boasted that it collected facts on every aspect of a person’s life, including “his marital troubles, jobs, school history, childhood, sex life, and political activities.” Unfortunately, much of that information was inaccurate and difficult to challenge (a problem that persists to this day). It was precisely these practices, coupled with the growing use of computerized databases for information processing, that led to the passage of the FCRA in 1970. (For more on the creation of the Fair Credit Reporting Act, see Arthur R. Miller, The Assault on Privacy: Computers, Data Banks, and Dossiers (Ann Arbor: University of Michigan Press, 1971).)
Investigations into personal habits were justified on the grounds that they revealed the character of the prospective borrower, which presumably shed light on his or her ability to repay the loan. The use of behavioral scoring is similarly premised on the notion that data aggregation can reveal one’s true character (at least with regards to financial stability) better than simpler methods such as one’s history with payments. While this can represent a great intrusion into personal information, it also raises the question of how exactly character is assessed.
In the older regime, character was essentially a proxy for how closely one approximated the middle-class white ideal. Assessments of character were used to screen out African-Americans, Jews, and the working class, among others. Behavioral screening avoids explicit appeals to any particular identity, but it has a similar effect. To the extent that a data profile tends to exhibit behavior associated with persons of low financial stability, its creditworthiness decreases.
There are two distinct problems with data aggregation as a window into character. The first is that the accuracy of profiles built on data aggregation is questionable. For example, you may have had the experience of having Amazon continue to recommend books based on a gift purchased for someone else. But concerns based on accuracy fail to get to the heart of the matter; accuracy is largely a technical problem. The deeper problem is that even with greater accuracy, data is necessarily incomplete. The data aggregators may argue that they are not making normative judgments at all, that they are just finding correlations among the data. But the result is the same. The challenge posed by the use of data aggregation is not only that information about ourselves circulates beyond our control and may not be interpreted in the right context, but rather the possibility that we will be able to make increasingly accurate assessments about our fellow citizens on the basis of the traces that they leave behind. The application of data aggregation to credit scoring is relatively simple, but it points to much larger things to come.