<p>The adverse impact of agricultural carbon emissions on climate change has drawn widespread attention around the world. Accurately identifying the impact of socio-economic factors on agricultural carbon emission intensity (ACI) is of vital importance for achieving the United Nations’ Sustainable Development Goals (SDGs) of reducing carbon emissions and addressing the climate crisis. In this paper, we collected 20 related explanatory variables across agricultural, economic, demographic, and governmental dimensions for 87 cities in central China from 2011 to 2020, and examined their impacts on ACI using machine learning models and interpretable methods. This study aims to identify the nonlinear effects of socio-economic variables on agricultural carbon emission intensity and reveal the interaction effects of several key variables on it. The results indicated that: (1) The XGBoost model achieved the best performance (<InlineEquation ID="IEq1"> <EquationSource Format="TEX">\(\:{R}^{2}=0.802\pm\:0.028\)</EquationSource> </InlineEquation>) on the testing data set. (2) Level of technology, multiple cropping index, food price index and irrigation index played a more critical role in driving ACI, while farmers’ income, per capita GDP and agricultural agglomeration might play a relatively weak role. (3) Irrigation index and multiple cropping index had sustained promotional and inhibitory effects on ACI, respectively. Moreover, variables such as urban-rural gap, food price index, per capita cultivated land, fixed investment in agriculture left a significant nonlinear effect on ACI. (4) There were differentiated interactive effects among multiple cropping index, irrigation index, food price index, urban-rural gap and level of technology on ACI. Our findings provide compelling evidence for understanding the mechanisms driving agricultural carbon emissions, which can be a reference for carbon mitigation in the agricultural sector.</p>