[guided]The point of this step is to isolate the finite-dimensional part of the argument. Algebraic closure in a strongly minimal theory behaves like linear span in a [vector space](/page/Vector%20Space): the exchange axiom is the replacement rule, and the statement says that an independent set contained in the closure of $n$ independent elements cannot have more than $n$ elements.
We first prove the replacement move carefully, because this is the place where exchange must be used with the correct hypotheses. Let $P \subseteq M$ be a parameter set, let $X=\{x_1,\dots,x_m\} \subseteq M$ be finite and independent over $P$, and let $y \in \operatorname{acl}(P \cup X) \setminus \operatorname{acl}(P)$. Since $X$ is finite, choose the least $r \in \{1,\dots,m\}$ such that
\begin{align*}
y \in \operatorname{acl}(P \cup \{x_1,\dots,x_r\}).
\end{align*}
The minimality of $r$ gives
\begin{align*}
y \notin \operatorname{acl}(P \cup \{x_1,\dots,x_{r-1}\}).
\end{align*}
Now the [exchange axiom for algebraic closure in a strongly minimal theory](/page/Pregeometry) applies over the parameter set $P \cup \{x_1,\dots,x_{r-1}\}$ to the two elements $y$ and $x_r$. It gives
\begin{align*}
x_r \in \operatorname{acl}(P \cup \{x_1,\dots,x_{r-1}\} \cup \{y\})
\subseteq
\operatorname{acl}(P \cup (X \setminus \{x_r\}) \cup \{y\}).
\end{align*}
Because $y \in \operatorname{acl}(P \cup X)$ as well, monotonicity and idempotence imply the closure equality
\begin{align*}
\operatorname{acl}(P \cup X)
=
\operatorname{acl}(P \cup (X \setminus \{x_r\}) \cup \{y\}).
\end{align*}
We also need the replacement set to be independent. If $(X \setminus \{x_r\}) \cup \{y\}$ were dependent over $P$, then, since $X \setminus \{x_r\}$ is already independent over $P$, the dependence must involve $y$. There are two possible forms. First, the dependence may be witnessed directly by $y$: for some finite $H \subseteq X \setminus \{x_r\}$,
\begin{align*}
y \in \operatorname{acl}(P \cup H).
\end{align*}
Second, the dependence may initially be witnessed by an element $z \in X \setminus \{x_r\}$. Then for some finite $H \subseteq X \setminus \{x_r,z\}$,
\begin{align*}
z \in \operatorname{acl}(P \cup H \cup \{y\}).
\end{align*}
Because $X$ is independent over $P$, the element $z$ is not in $\operatorname{acl}(P \cup H)$. Therefore the exchange axiom applies over the parameter set $P \cup H$ to $z$ and $y$, and gives
\begin{align*}
y \in \operatorname{acl}(P \cup H \cup \{z\}) \subseteq \operatorname{acl}(P \cup (X \setminus \{x_r\})).
\end{align*}
Thus in either form of dependence, we obtain $y \in \operatorname{acl}(P \cup (X \setminus \{x_r\}))$. Substituting this into the relation placing $x_r$ in the closure of the previous elements and $y$ gives
\begin{align*}
x_r \in \operatorname{acl}(P \cup (X \setminus \{x_r\})),
\end{align*}
by monotonicity and idempotence. This contradicts the independence of $X$ over $P$. Hence $(X \setminus \{x_r\}) \cup \{y\}$ is independent over $P$.
Now let $E,F \subseteq M$ be finite, assume both are independent over $P$, and assume
\begin{align*}
E \subseteq \operatorname{acl}(P \cup F).
\end{align*}
We prove $|E| \leq |F|$ by induction on $|E|$, uniformly in the parameter set $P$. The uniform formulation is needed because after replacing one element of $F$ by one element of $E$, the induction is applied over the enlarged parameter set $P \cup \{e\}$.
If $E=\varnothing$, then $|E|=0 \leq |F|$. Assume $E$ is nonempty. Choose $e \in E$ and define
\begin{align*}
E_0 := E \setminus \{e\}.
\end{align*}
Since $E$ is independent over $P$, we have $e \notin \operatorname{acl}(P)$. Since $E \subseteq \operatorname{acl}(P \cup F)$, we have $e \in \operatorname{acl}(P \cup F)$. Applying the replacement lemma to $X=F$ and $y=e$ gives an element $f \in F$ such that
\begin{align*}
\operatorname{acl}(P \cup F)
=
\operatorname{acl}(P \cup (F \setminus \{f\}) \cup \{e\})
\end{align*}
and $(F \setminus \{f\}) \cup \{e\}$ is independent over $P$. This independence exactly says that $F \setminus \{f\}$ remains independent after $e$ is moved into the parameter set, that is, $F \setminus \{f\}$ is independent over $P \cup \{e\}$.
The set $E_0$ is independent over $P \cup \{e\}$ because $E$ was independent over $P$. The closure equality gives
\begin{align*}
E_0 \subseteq \operatorname{acl}(P \cup F)
=
\operatorname{acl}(P \cup \{e\} \cup (F \setminus \{f\})).
\end{align*}
Therefore the induction hypothesis applies over $P \cup \{e\}$ to $E_0$ and $F \setminus \{f\}$, yielding
\begin{align*}
|E|-1 = |E_0| \leq |F \setminus \{f\}| = |F|-1.
\end{align*}
Adding $1$ to both sides gives $|E| \leq |F|$.[/guided]