My book states the following theorem:
Suppose $K$ and $L$ are fields and $i:K\to L$ an isomorphism between them. Let $K(\alpha)/K$ and $L(\beta)/L$ be simple algebraic extensions, such that $\alpha$ has minimum polynomial $m_\alpha$ over $K$, and $\beta$ has minimal polynomial $m_\beta$ over $L$. Suppose further that $i(m_\alpha(t))=m_\beta(t)$. Then there exists an isomorphism $j:K(\alpha)\to L(\beta)$ such that $j(k)=i(k)$ for all $k\in K$, and $j(\alpha)=\beta$.
Is it true, in general, that if two simple algebraic field extensions are isomorphic, then such an $I$ needn't exist? Looking at $K=L=\mathbb{Q}$, $m_\alpha=t^2-2$ and $m_\beta=t^2-4t+2$, the extensions are isomorphic, but there is no automorphism $i:\mathbb{Q}\to \mathbb{Q}$ such that $i(m_\alpha(t))=m_\beta(t)$.