This dissertation proceeds in two steps, a theoretical one followed by an empirical one. Considered together, they help us understand how persuasive language may be used to endorse or justify scientifically shaky policies.In the first step, I introduce and define a troubling phenomenon that may occur in public decision making, Audacious Use of an Information Tool (AUIT). AUIT is curious because it involves a decision not to defer to the scientific community on a question that solicits their expertise: whether a tool of science is methodologically capable doing what it is being asked to do. AUIT is potentially dangerous because it involves high-stakes decisions or plans based on possibly faulty information. I identify two cases of AUIT: the use of value-added modeling (VAM) in teacher evaluation and the use of predictive genetic testing to make personal health-care decisions.In the second step, I map the language of AUIT using a case-study design. I focus on the first example mentioned above, the use of VAM as a central component in high-stakes personnel decisions about individual teachers. My rhetorical analysis considers a corpus of texts from 2008 to 2015, drawn from popular media, advocacy discourse, and policy messaging from the Department of Education. I identify four rhetorical moves that rationalize the Audacious Use of VAM in teacher evaluation and conjecture about whether they are likely to be deployed in other cases of AUIT.