One of the most striking features of human intelligence is our capacity to rapidly and flexibly plan. Planning enables us to solve myriad everyday problems---e.g., planning how to complete a list of errands on a busy day---but planning is also very computationally demanding. How do we effectively plan despite fundamental constraints on our time, memory, and attention? My talk will cover recent work investigating how the flexible construction of task representations facilitates efficient planning in humans. I will discuss value-guided construal, a general formal theory of how people form simplified, ad hoc representations in order to plan and act. By investigating the computational principles that underlie how people construe problems, this approach provides a new perspective on the dynamic interplay of attention, causal reasoning, and goal-directed behavior.