Tay explores the Performance-Interpretability Trade-off (PIT) as a critical tension in AI and machine learning, and shows its distinctive form in discourse analysis where predictive success and interpretive meaning are inseparable.
Rather than treating PIT as a technical obstacle, this book reframes it as a site of conceptual negotiation and theoretical innovation. It introduces constructs such as strategic indeterminacy and PIT elasticity alongside analytic strategies like discourse fingerprinting, to show how discourse knowledge can actively reshape computational assumptions at every level of the analytic pipeline. Through sustained case studies, the book equips readers to engage machine learning algorithms as a partner in interpretation and methodological reflection.
An essential resource for scholars and researchers in linguistics, discourse analysts, computational linguists, and digital humanities, offering a comprehensive roadmap for harnessing machine learning's transformative potential.