Using a state-of-the-art semi analytic model for galaxy formation, we investigated in detail the effects of black hole (BH) accretion triggered by disk instabilities (DI) in isolated galaxies on the evolution of the AGN population. Specifically, we took on, developed, and expanded the Hopkins & Quataert (2011, MNRAS, 411, 1027 model for the mass inflow following disk perturbations, based on a physical description of nuclear inflows and tested against aimed N-body simulations. We compared the evolution of AGN due to such a DI accretion mode with that arising in a scenario where galaxy interactions (IT mode) produce the sudden destabilization of large quantities of gas feeding the AGN; this constitutes the standard AGN feeding mode implemented in the earliest versions of most semi-analytic models. To study the maximal contribution of DI to the evolution of the AGN population, we extended and developed the DI model to assess the effects of changing the assumed disk surface density profile, and to obtain lower limits for the nuclear star formation rates associated to the DI accretion mode. We obtained the following results: i) For AGN with luminosity M 1450 > ∼ −26, the DI mode can provide the BH accretion needed to match the observed AGN luminosity functions up to z ≈ 4.5. In such a luminosity range and redshift, it constitutes a viable candidate mechanism to fuel AGN, and can compete with the IT scenario as the main driver of cosmological evolution of the AGN population. ii) The DI scenario cannot provide the observed abundance of high-luminosity QSO with M 1450 ≤ −26 AGN, as well as the abundance of high-redhshift z > ∼ 4.5 QSO with M 1450 ≤ −24. As found in our earliest works, the IT scenario provides an acceptable match to the observed luminosity functions up to z ≈ 6. iii) The dispersion of the distributions of Eddington ratio λ for low-and intermediate-luminosity AGN (bolometric L AGN = 10 43 −10 45 erg s −1 ) is predicted to be much smaller in the DI scenario compared to the IT mode. iv) The above conclusions concerning the DI mode are robust with respect to the explored variants of the DI model. We discuss the physical origin of our findings. Finally, we discuss how it is possible to pin down the dominant fueling mechanism of AGN in the low-intermediate luminosity range M 1450 > ∼ −26 where the DI and the IT modes are both viable candidates as the main drivers of the AGN evolution. We show that an interesting discriminant could be provided by the fraction of AGN with high Eddington ratios λ ≥ 0.5, since it increases with luminosity in the IT case, while the opposite is true in the DI scenario.