This review examines the evolving role of representations—neural, mental and computational constructs that mediate perception, reasoning, and behavior—across the intertwined histories of cybernetics, cognitive science, and artificial intelligence. We begin by revisiting the rise and decline of cybernetics and its influence on the cognitive revolution, foregrounding how early debates over internal representations continue to shape current approaches. We then explore three frameworks that exemplify how representations function in scientific practice: cognitive maps in neuroscience, schemas in cognitive and clinical psychology, and object relations in psychodynamic theory. These traditions, though methodologically distinct, converge in treating representations as adaptive interfaces tailored to disciplinary goals rather than static mirrors of reality. Drawing on insights from computational psychiatry and neuroAI, we argue that scientific progress depends not on converging toward a single “correct” model, but on developing representations that are flexible, context-sensitive, and reflexively engaged with their own assumptions. We caution against equating engineering performance with theoretical adequacy and highlight how historical cycles of scientific hype, from cybernetics to modern AI, can obscure valuable, if marginalized, contributions. By integrating perspectives across psychology, neuroscience, and AI, this review offers conceptual tools to support transdisciplinary dialogue and encourages researchers to treat representation not as a fixed ontology, but as a problem space requiring ongoing negotiation and pluralistic experimentation.