Gravar-mail: Learning to Overexert Cognitive Control in a Stroop Task