In this paper we adopt a generalized recursive smooth ambiguity model which permits a three-way separation among risk aversion, ambiguity aversion, and intertemporal substitution. We specify the representative agent's learning process and state beliefs, and apply this utility model to a consumption-based asset pricing model. When the dividends follow hidden Markov regime-switching processes, and the representative agent has ambiguous beliefs about the hidden state, we derive the implication of a