Railsback & Grimm wrote an excellent introduction to agent-based models. It is quite accessible and can easily be used to teach ABM in the social sciences. I did miss some of the bits that make the current standard choice in the social sciences (Gilbert & Troitzsch) such a good book, like on the nature of modelling in general, or on the advantages of ABM. In this sense, chapter 19 in R&G should come much earlier in the book. Jumping in with an example, as they do, has some advantages, but in my view they’re holding back too much.
Interestingly, at the very end of the book, they seem to undersell ABM. Not that I expect ABM to revolutionize the way social sciences is done, but they don’t mention the advantages for students to learn ABM: While everyone talks mechanism these days, ABM forces students to go beyond mere talk.