Time allocation is an irreversible process, which cannot be undone if it is found to be suboptimal.
When learners lack the information or cognitive capacity to compute optimal allocations ex-ante,
they are often forced to rely on local cues to adjust their decisions during the process of allocation
itself. In certain cases, such as concave learning curves that are independent across items, there
exists a simple adjustment rule that ensures optimal terminal allocations. Even in such cases,
however, precise local knowledge of empirical learning curves is necessary in order for adaptive
learning to result in optimal allocation. With certain other commonly observed learning curves such
as the logistic, achieving optimality through adaptive adjustments is an even greater challenge, as
learning based on local cues can result in suboptimal terminal allocations at least for some values
of the time budget.