More dynamic map elements is the obvious addition. We had some experimentation with this in Halo 2 and 3, but from there it kind of went downhill. Like anything, it has the potential to be nothing but a huge gimmick if executed purely, but it’s an aspect of map design that has been explored far too little. Many interesting things could be achieved by both player-activated elements, and segments of the map that move on their own.
However, I stress that these dynamic elements should rather come in the form of deliberate set pieces rather than some kind of generic system such as fully destructible environments. The reason is that individual pieces are something the level design has much more control over, and can tailor them to create a very special unique style for a given map. In contrast, something like destructible environments is too generic. It’s difficult to control and tailor for a specific experience, and is essentially the same on every map, and therefore it has less potential for creativity and truly interesting applications.
Another idea that is kind of interesting are maps where the direction of gravity is not constant à la Super Mario Galaxy. For a shooter such a map is very much a double edged sword, because it’s very easy to disorient players, especially regarding the direction from which they are being shot at. However, it is again something that has potential for interesting designs. It also has interesting meta-implications regarding the interaction of adjacent gravity wells with different orientations, and how that can be used for movement. On the other hand, this means more potential to seriously break maps, which puts more burden on playtesting (and more restrictions on design).
The final idea that could be interesting to explore are Portal like doorways that seamlessly connect two completely separate parts of a map. This would allow creation of maps that are topologically impossible with standard map design, which obviously has huge gameplay implications. Creation of such maps would require incredible intuition for topology, so this too has the change of going horribly wrong.
Each of these features would require putting significant effort into creating a robust system that allows the level designers to easily use the feature. As a consequence, it would be something the team would have to commit to and not use on just one or two, but on a number of maps, which amplifies the risk of poorly designed maps due to unfamiliarity with the features and tools. Dynamic map elements are probably the safest one, because they are smallest departure from traditional design, and can be very varied in both style and scale. The last two ideas are more dreamy, risky, and potentially don’t even have any greater pay-off. But as far as ideas go, I think they help to demonstrate how much room for evolution and innovation there is left in map design if developers actually gave any attention to it.