For $i\in\{1,2,3,4\}$, let nonnegative integer decision variable $x_i$ be the number of tables occupied by exactly $i$ people. The problem is to maximize $x_1$ subject to linear constraints
\begin{align}
x_1 + x_2 + x_3 + x_4 &= 16 \tag1\label1 \\
1 x_1 + 2 x_2 + 3 x_3 + 4 x_4 &= 50 \tag2\label2
\end{align}
The linear programming relaxation yields optimal solution $x^*=(14/3,0,0,34/3)$, and the LP dual solution $(4/3,-1/3)$ certifies optimality:
\begin{align}
x_1 &\le x_1 + \frac{2}{3}x_2 + \frac{1}{3}x_3 \\
&= \frac{4}{3}(x_1 + x_2 + x_3 + x_4)
-\frac{1}{3}(1 x_1 + 2 x_2 + 3 x_3 + 4 x_4) \\
&= \frac{4}{3}\cdot16-\frac{1}{3}\cdot50 \\
&= \frac{14}{3}
\end{align}
So the ILP has optimal objective value at most $\lfloor 14/3 \rfloor = 4$. Now find such a solution $x=(4,\underline{},\underline{},\underline{})$.
Note that the optimal LP solution can reasonably be obtained by inspection by arguing that $x_1$ will be maximized when $x_2=x_3=0$, which reduces \eqref{1} and \eqref{2} to two equations and two unknowns.
An alternative approach skips the LP solve and instead uses “probing” to deduce the upper bound of $4$. Suppose there are at least five tables with exactly one person ($x_1\ge5$). Then we must distribute the remaining $50-5=45$ people among the remaining $16-5=11$ tables that together have capacity $44<45$, but that is impossible. So $x_1 \le 4$.