The event horizon in an asymptotically flat spacetime (which is not quite the universe we live in, but a decent first step) is defined as the causal past of the infinite causal future. This definition guarantees that we see no effects whatsoever from the part of the universe that is behind the event horizon. The problem with this definition is that we have to wait forever to draw the horizon accurately. Thus there are several alternative horizons which are more instrumentally useful for theorem proving and/or numerical simulations, but are not in general identical to the event horizon. The cosmological event horizon is a totally different beast (it is similar to the Rindler horizon, used to derive the Unruh effect), though it does share a number of properties with the black hole event horizon. There are further exciting complications once you get deeper into the subject.
The event horizon in an asymptotically flat spacetime (which is not quite the universe we live in, but a decent first step) is defined as the causal past of the infinite causal future. This definition guarantees that we see no effects whatsoever from the part of the universe that is behind the event horizon. The problem with this definition is that we have to wait forever to draw the horizon accurately. Thus there are several alternative horizons which are more instrumentally useful for theorem proving and/or numerical simulations, but are not in general identical to the event horizon. The cosmological event horizon is a totally different beast (it is similar to the Rindler horizon, used to derive the Unruh effect), though it does share a number of properties with the black hole event horizon. There are further exciting complications once you get deeper into the subject.