[Coda]

Saying "force holding it together" was a handwave on my part. It doesn't actually work that way but it's close enough to give you the general idea. An object's strength (and there are several different kinds of strength to measure various different kinds of interactions) is measured in force per unit area (that is, pressure), and it measures how much pressure can be applied to the object before it undergoes a permanent deformation (such as bending or breaking) that it can't bounce back from. So air bags actually provide an ADDITIONAL benefit beyond just spreading out the impulse over time -- it ALSO spreads out the force over a larger surface area compared to hitting your steering column chest-first, meaning there's less pressure being applied to any one part of your body.

(Crumple zones in the structure of a car also protect the occupants by increasing the impulse time, since the cab of the car ends up decelerating more slowly than the point of contact. They don't do anything about increasing the surface area of the contact against your body like an air bag does, although they do additionally reduce the force by using up some of the kinetic energy by deforming -- after all, it takes energy to make the metal bend.)

I wouldn't describe work as one interaction. But if you're considering the end result of a series of actions and figuring out the work done by that, then you can add up the force vectors over time and get the same answer as if the integrated force was the only thing that happened. It's analogous to our earlier discussion of average velocity -- you're choosing to disregard the in-between details to come up with an overall description of what happened in the long run.

And yes, you're basically right about net force. To get into a technicality of the math, the operation isn't actually ADDING the vectors, but INTEGRATING them over time, because you could balance out a force of 2N over 1s with a force of 1N over 2s. As long as you're working with nice round-number quantities like that, you can get the right result by inspection (perhaps breaking up the 1N/2s impulse into two 1N/1s impulses so that you've got consistent time units) but generalizing it beyond that gets into calculus.