Let $F$ be a compact oriented surface and $x_0\in F$ a basepoint. Consider the set $\mathcal E=Emb_0(I,F\times I)$ of embeddings $\sigma\colon I\hookrightarrow F\times I$, $\sigma(\partial I)=\{x_0\}\times\partial I$, isotopic with fixed ends to the trivial long knot $\{x_0\}\times I$.
The natural projection $p\colon F\times I\to F$ maps a string $\sigma\in\mathcal E$ to a contractible loop $p\circ\sigma$ in $F$. On the other hand, the graphics $\Gamma_\omega\subset F\times I$ of a contractible loop $\omega\in\Omega(F,x_0)$ is isotopic to the trivial string. Thus, the space of contractible loops $\Omega_0 (F,x_0)$ is a retract of the space of long unknots $\mathcal E$.
My question is whether $\Omega_0 (F,x_0)$ is homotopy equivalent to $\mathcal E$.
In particular, this would imply that $\pi_i(\mathcal E)=\pi_i(\Omega_0 (F,x_0))=\pi_{i+1}(F)$, $i\ge 1$. For example, when $F=S^2$, $\pi_2(\mathcal E)=\pi_{3}(S^2)=\mathbb Z$, and the space of long unknots in the sphere is not trivial.
When $F$ is a disk, the answer is positive: by A. Hatcher the space of long unknots is contractible.