My answer is based on story points as estimation of size & complexity. When estimating points, it's useful to remember that it's like estimating distance -- if you're following a path, that path is n miles long no matter if you're walking, cycling, or driving a car. If you're estimating the complexity of a story, it should be the same no matter if Joe is doing it or Jane is doing it or Bob is doing it.
The actual time it takes to get to your destination (or complete the story) depends on the vehicle; a car will get you to your destination faster than walking will, much like one person may complete the story a bit quicker than someone else. Here's where the analogy breaks down, though, because as pdr says in this comment, "learning curves should not be included in an estimate; we should assume that we both have each other available to teach." And you should assume that anyone could pick up the story and complete it, so the estimation should be one that the group determines together if you want accurate determination of velocity and so forth.
It's worth bearing in mind that if you estimate points based on a 1:1 relationship between individual and story, the rest of your stories should be estimated as such; similarly, if you estimate stories based on multiple people working on them at once, all of them should be like that, otherwise your determination of size and complexity would be off and any calculation of metrics you might use would be relatively meaningless (e.g. what does it mean if you have 5 3 point stories but 3 of them are 3s if everyone works on them and 2 of them are 3s if they're worked on individually? it means you probably want to be measuring the same things all the time.)