The answer lies in the derivation of the VIX, as implemented by the CBOE.
The basic derivation of the VIX was done by Demeterfi et al. (1999), where they used a "basket" of options to replicate expected future variance. This yields the formula:
$$\begin{aligned}
\mathbb{E}[\mathbb{V}] =& \frac{2}{T} \left[ rT - \left( \frac{S_0 e^{rt}}{S_\star} - 1 \right) - \ln\left(\frac{S_\star}{S_0} \right) \right] \\
&+ e^{rT} \int_0^{S_\star} \frac{1}{K^2} P_0(K) dK\\
&+ e^{rT} \int_{S_\star}^\infty \frac{1}{K^2} C_0(K) dK\\
\end{aligned}$$
With risk-free-rate $r$, time to expiration $T$, $S_0$ the initial stock price, $S_\star$ a boundary price and $P$ and $C$ representing put and call options with strike price $K$ respectively.
The CBOE then approximates the first line by
$$
-\frac{1}{T} \left( \frac{F_0}{K_0} - 1 \right)^2
$$
as shown by Jiang and Tian (2007). Using numerical integration, the integrals turn into the sums seen in the CBOE formula.
So in a purely technical way the answer is that you use the forward price in order to get rid of the first term, making the calculation feasible.
For a full derivation of the CBOE VIX with steps you can look up Appendix A in my paper No Model No Cry? on SSRN.