Updating Beliefs when Evidence is Open to Interpretation: Implications for Bias and Polarization