Roy writes about what to do when people don't want to do pair programming. The gist of his post is how do you convince people into pair programming?

My opinion is that if you force people to do something, you'll never get as much productivity as when they want to do something. That’s the whole point about getting developers to choose the work they want to do rather than assign it, to have them give estimates rather than assign deadlines. Think about it. The last time someone told you a deadline, were you all that motivated to achieve it on time? Probably not. But when someone asks for your honest estimate of the date, I’ll bet you worked whatever time necessary to make it unless your estimate was horribly off, which is a common problem in software estimation.

I'm not all that big on pair programming. I think a combo of pair design and code reviews results in equivalent quality and happier programmers. That’s not to say I encourage excessive up-front design, but the most value that can be added to software is through a good design. Good design skills separate the programmers from the architects.