You could trivially prompt it in, asking the model to never be persuaded. But that works against the final goal of collective decision making. Finding a generalizable trade-off between susceptibility and resistance is not trivial given the large cross-model behavioral variation.
