In this paper, we propose an approach to describe the behavior of naive agents with quasi-hyperbolic discounting in the neoclassical growth model. To study time-inconsistent decision making of an agent who cannot commit to future actions, we introduce the notion of sliding equilibrium and distinguish between pseudo-perfect foresight and perfect foresight. The agent with pseudo-perfect foresight revises both the consumption path and expectations about prices; the agent with perfect foresight correctly foresees prices in a sliding equilibrium and is naive only about their time inconsistency. We prove the existence of sliding equilibria for the class of isoelastic utility functions and show that generically consumption paths are not the same under quasi-hyperbolic and exponential discounting. Observational equivalence only holds in the well-known cases of a constant interest rate or logarithmic utility. Our results suggest that perfect foresight implies a higher long-run capital stock and consumption level than pseudo-perfect foresight.