The Effect of Emerging Technologies on Democratic Governance

• In Event: (Mis)Information in Great Power Competition and Democratic Governance

Abstract

Advances in machine learning have created natural language models that can mimic human writing style and substance. The technology can increase the efficiency of journalism but the ability to generate synthetic, yet not plagiarized text may also present challenges to democracy. While responsiveness to constituency preferences is both electorally wise and a definition of sound governance, using technology to manufacture inauthentic constituency positions is tantamount to unequal participation and influence. In this research, we conduct a field experiment in which we send both hand-written and machine-generated letters—across six policy issues and on the right and the left—to ~7200 state legislators to test the potential for advances in machine learning and text prediction to distort the process of democratic representation. We compare legislative response rates and tone for the authentic versus machine-generated constituency letters to gauge whether emerging technology can both approximate and scale up inauthentic constituency participation and manipulate democratic representation.

Author

• Sarah E. Kreps