Douglas Merrill of ZestFinance is quoted in the article Beyond Mere Numbers, on new measures being used by credit scorers, as saying he finds the use of social media data to determine creditworthiness “personally creepy.” He ought to be thinking harder about the metrics his firm is deciding to use instead.
The use of borrowers’ writing style, like using all capital letters, strikes me as presenting a risk of compounding other disadvantages that borrowers may be facing and have no control over, like their early educational history. Such data use could potentially even develop into new methods of red-lining.
Scholars of linguistics, like William Labov, have described the grammatical differences among different ethnic and regional groups in the United States, and have demonstrated that rather than being “bad English grammar,” they have their own grammatical rules. More recently, linguists have studied patterns in online writing. While gatekeepers, like educators and college admissions counselors, may judge these forms of writing as a strike against the writer, they may be unaware of the other purposes such writing styles have in people’s lives, like communicating emotion or belongingness in a particular community. Linguistic bias is one of the last unquestioned biases people can openly hold in the United States without being called out for discrimination.
Is including all-caps writing double-counting another measure in Merrill’s score, like educational achievement? Is it serving as a stand-in for race? And what other linguistic data is ZestFinance including in their score? We don’t know, but I think all of these factors and their impacts on borrowers should be made clear.
Ironically, if firms like ZestFinance and SoFi hired outside of “entrepreneurs with backgrounds in finance, software, and business consulting,” and overcame their stated biases against people who majored in things like Romance languages, they might have addressed these concerns already.