It's a slippery slope.
A simple Transfer object (or Parameter object) may have the sole purpose of holding some fields and providing their values on demand. However, even in that degenerate case one could argue that the object should be immutable -- configured in the constructor and exposing only get... methods.
There's also the case of a class that exposes some "control knobs"; your car radio's UI probably can be understood as exposing something like getVolume, setVolume, getChannel, and setChannel, but its real functionality is receiving signals and emitting sound. But those knobs don't expose much implementation detail; you don't know from those interface features whether the radio is transistors, mostly-software, or vacuum tubes.
The more you begin to think of an object as an active participant in a problem-domain task, the more you'll think in terms of asking it to do something instead of asking it to tell you about its internal state, or asking it for its data so other code can do something with those values.
So... "evil"? Not really. But every time you're inclined to put in a value and expose both get... and set... methods on that value, ask yourself why, and what that object's reponsibility really is. If the only answer you can give yourself is, "To hold this value for me", then maybe something besides OO is going on here.
