A dogwhistle is a type of expression used by politicians to send a signal of ideolo- gical affinity to one part of the population without being fully understood by another part of the population who may be alie- nated by the message. For example, a classic dogwhistle in US politics is ”inner city crime”, which had been used to send a signal of affinity with racist voters while simultaneously maintaining plausible de- niability through concern about serious social problems. A dogwhistle works only if some of the population ”gets it” and fails if everyone understands it well eno- ugh to act on it: ”adversarial vagueness”. In our new project, the Gothenburg Re- search Initiative for Politically Emergent Systems (GRIPES), we hypothesize that this property makes dogwhistles ”ga- meable” through computational methods, although organizations with an interest in doing so also have an interest in ensuring that their use of such methods remains obscure. This talk discusses GRIPES’ plans for an interdisciplinary approach to analyzing the limits of dogwhistle-style linguistic manipulation through computa- tional methods such as agent-based simulation.