First of all, thanks for taking serious my suggestion as put forward in the thread you mention. Two comments, though:
1a.) Instead of an identity matrix, I ended up thinking more of an identity map, compare the PPS in the reply Very interesting, implemented simply as I:x -> x (here, I stands for Identity, as I do not presently know how to type the 'double bar identity' symbol). This map is algebraic too, and as far as I can tell, its contents is not displayed either. With such a map, I(psi) would immediately display as psi, while the Dirac anticommutator algebra would still display as 2 times the metric times this I. But then again, there might be a problem: The anticommutator acting on a spinor would, I guess, have to entered as
perhaps unnatural, rather than
1b.) Having such an identity map instead of the identity matrix, the latter implemented using the deprecated (right?) matrix constructor, would, I think, also be a conceptually cleaner solution.
2.) I think it would be more natural/intuitive if this identity element was not to be appended manually, as it is done in Eq. (6), but was present from the outset in Eq. (1).
PS: How is the 'double bar identity' symbol actually entered in Maple?