Let $\operatorname{li} x$ denote the logarithmic integral $$\operatorname{li} x=\int_0^x\frac{dt}{\ln t}.$$ Consider the following parameterized integral: $$I(a)=\int_0^1\left(-\frac{\operatorname{li} x}x\right)^adx.$$
Can we find a closed form for this integral?
We can find some special values of this integral: $$I(0)=1,\,\,I(1)=1,\,\,I(2)=\frac{\pi^2}6,\,\,I(3)\stackrel?=\frac{7\zeta(3)}2$$ The last value was suggested by numeric computations, and I do not yet have a proof for it.
Can we prove the conjectured value of $I(3)$?
One could expect that $I(4)$ might be a simple rational (or at least algebraic) multiple of $\pi^4$ but I could not find such a form.
Can we find closed forms for $I(4),I(5)$ and other small integer arguments?
$I(5)$ doesn't seem to have a closed form that I could find. On the other hand, I found numerically that $$ \begin{eqnarray}I(4) &=& -\tfrac{32}{3} \text{Li}_4(\tfrac{1}{2})+\tfrac{52}{3} \zeta(4)+\tfrac{8}{3} \zeta (2) \log^22-\tfrac{4}{9}\log^4 2. \end{eqnarray}$$
Case 3. The way to do the integral $I(3)$ is to write $$ \mathrm{li}(x) = -E_1(\log\tfrac1x), $$ and make the change of variable $t=1/u$, so that $$ I(a) = \int_1^\infty u^{a-2}E_1(\log u)^a\,du. $$ This way, the exponential integral takes the form $$ E_1(x) = \int_1^\infty \frac{dt}{t} e^{-x t}, $$ so that the integral $I(a)$ is equal to $$ I(a) = \int_1^\infty \frac{u^{a-2-\sum t}}{t_1\cdots t_a}\,du\,dt_1\cdots dt_a. $$ The integral over $u$ can be done easily enough: $$ I(a) = \int_1^\infty \left(\prod \frac{dt}{t}\right) \frac{1}{\sum t+1-a}. $$
For $a=3$, Mathematica can do this three-dimensional integral: $$ I(3) = \int_1^\infty\frac{ds\,du\,dv}{suv}\frac{1}{s+u+v-2} = \tfrac72\zeta(3). $$