Early AI planning systems borrowed key ideas from studies of human problem solving, but the past
30 years have seen researchers largely abandon this strategy. This has led to computational methods
that are highly efficient for certain classes of problems, but that lack many desirable features of
people’s cognition. In this paper, I review the initial interactions between the two fields, the reasons
for their gradual separation, and the potential benefits of renewed interchange. I examine abilities
that human problem solvers exhibit but that receive scant attention in the planning literature, along
with ideas for addressing these omissions. In closing, I suggest ways to encourage more research
in AI’s original interdisciplinary tradition, which the cognitive systems paradigm adopts.