Here's a imaginary dialog of one way they could choose to play the game:
$\ $Dan: | What's the set of vertices? Let's call it $\mathsf \C_0$, the $0$-dimensional cells. |
$\ $Eliza: | I'll decide $\C_0$ is the set $\{A, B, C\}$. We want there to be morphisms, $1$-dimensional cells,
in our structure, but I can't decide what they are until I know what their boundary is allowed to be in terms of the $\C_0$ I just decided on.
Let's call this $\B_0$, the $0$-dimensional boundaries. What's in that set? |
$\ $Dan: | I'll decide that $\B_0 = \C_0 \x \C_0$. Every $1$-cell will need a domain and codomain.
What's the set $\C_1$ of morphisms, then? I also want to know how to assign boundaries to all of them: I'll need a function $\partial : \C_1 \to \B_0$ |
$\ $Eliza: | I'll decide $\C_1$ is the set $\{f, g, h\}$, and I'll say that
$\partial f = \< A, B \>$ and $\partial g = \< B,C \>$ and $\partial h = \< A, C\>$. Now what's the set $\B_1$ of 1-dimensional shapes
that serve as boundaries for $2$-cells? I'll also need to know $\partial : \B_1 \to \B_0$ to know how these boundaries relate to lower-dimensional
ones. Hope you don't mind that I'm overloading $\partial$. |
$\ $Dan: | No
problem. As for $\B_1$, let me introduce an auxiliary concept.
Say $\P_1$ is the set of paths over $\C_1$. That is, an
element of $\P_1$ is a list $[f_1, \ldots, f_n]$ of elements of
$\C_1$, such that the codomain of $f_i$ (that is, $\pi_2 \partial
f_i$) is equal to the domain of $f_{i+1}$ (that is, $\pi_1 \partial f_{i+1}$). And I can define
$\partial : \P_1 \to \B_0$ by saying that the domain of $[f_1, \ldots, f_n]$ is the
domain of $f_1$, and its codomain is that of $f_n$.
Now I can say that $\B_1$ is a pair of paths in $\P_1$, with the same boundary.
That means $\partial :\B_1 : \B_0$ has an obvious definition: just take either path, and return its boundary.
Now what's $\C_2$ and $\partial : \C_2 \to \B_1$?
|
$\ $Eliza: | Ok, I'll decide $\C_2$ is the set $\{\alpha, \beta\}$, and I'll say that
$\partial \alpha = \< [f,g], h \>$ and $\partial \beta = \< h,[f,g] \>$. Now what's the set $\B_2$ of 2-dimensional shapes
that serve as boundaries for $3$-cells, and what's $\partial : \B_2 \to \B_1$?
|
$\ $Dan: | To define $\B_2$, let me again introduce a notion of "path".
Say $\P_2$ is the set of pasting diagrams made up of cells from $\C_2$. [...]
|
$E_0$) | $\ $Eliza | chooses a set $\C_0$. |
$D_0$) | $\ $Dan | chooses a set $\B_0$. |
$E_1$) | $\ $Eliza | chooses a set $\C_1$, and a map $\partial : \C_1 \to \B_0$. |
$D_1$) | $\ $Dan | chooses a set $\B_1$, and a map $\partial : \B_1 \to \B_0$. |
$E_2$) | $\ $Eliza | chooses a set $\C_2$, and a map $\partial : \C_2 \to \B_1$. |
$D_2$) | $\ $Dan | chooses a set $\B_2$, and a map $\partial : \B_2 \to \B_1$. |
$\vdots$ |
$E_n)$ | $\ $Eliza | chooses a set $\C_n$, and a map $\partial : \C_n \to \B_{n-1}$. |
$D_n)$ | $\ $Dan | chooses a set $\B_n$, and a map $\partial : \B_n \to \B_{n-1}$. |
(For uniformity we could set $\B_{-1} = \top$, and have Dan and Eliza choose degenerate maps into $\top$ on turns $E_0$ and $D_0$)
A morphism from Eliza's "graph" to Eve's should surely at least involve maps of cells $f_i : \C_i \to \C'_i$,
and we'd like to be able to assert that boundaries are preserved. But problematically, the boundaries
don't have the same types:
\[\begin{CD}
\C_1 @>\partial>> \B_0 \\
@V{f_1}VV @VV{???}V\\
\C'_1 @>>{\partial}> \B'_0
\end{CD}\]
I think we need to assume that Dan's strategy is suitably functorial, so that we can write
something like
\[\begin{CD}
\C_1 @>\partial>> \beta_0(C_0) \\
@V{f_1}VV @VV{\beta_0(f_0)}V\\
\C'_1 @>>{\partial}> \beta_0(C'_0)
\end{CD}\]
with $\beta_0$ being a functor $\rset \to \rset$ such that $\B_0 = \beta(C_0)$ and $\B'_0 = \beta(C'_0)$.