In climate model experiments conducted at various carbon dioxide (CO2) concentrations, we found that arctic amplification (AA) under cooling conditions is stronger than under warming conditions, and that the seasonality of AA differs between the two scenarios. Feedback analysis revealed that Planck, lapse-rate, and albedo feedbacks are important factors. Further analysis of the seasonality of each feedback and energy source discussed the intensity and causes of the differential strength of AA in cooling and warming scenarios. This analysis provides insight into how carbon reduction can affect Arctic climate, ecosystems, and socio-economic activities.