Abstract:

The evolution of smart agriculture has revealed a significant "large-farm bias," leaving 500 million smallholders worldwide at risk of systematic marginalization within algorithmic iterations. Exposing these structural inequalities and constructing inclusive empowerment mechanisms have become core imperatives for the development of digital villages. Integrating perspectives from human-computer interaction (HCI) psychology and Science and Technology Studies (STS), this research systematically reviews 33 core publications. By employing qualitative analysis and logical modeling, it constructs an "Agricultural AI Exclusion Cycle" framework to deconstruct the deep-seated mechanisms of data dividend aggregation and exclusion within the context of China’s "large country with small-scale farmers". The results indicate that: (1) Digital transformation has undergone a qualitative mutation into "data poverty," fundamentally reducing farmers’ status from "data producers" to "data objects"; (2) In the Chinese context, data poverty is characterized by non-standardized collection, benefit inversion, and a crisis of subjectivity; (3) The study deconstructs a "double helix" misalignment mechanism, where large-scale farms secure an "algorithmic premium" through a positive spiral, while smallholders fall into a negative exclusion spiral due to being treated as "data noise"; (4) Current governance is challenged by elite capture under "algorithmic authority" and the risk of "digital suspension". This study proposes that smart agriculture must pivot from "efficiency fetishism" toward "inclusive empowerment". At the policy level, fairness-based evaluations should be integrated, and the establishment of village-led "farmer data cooperatives" should be explored. At the technical level, edge-side "small-but-beautiful" AI models adapted to fragmented environments should be developed based on "trusted data spaces." This research provides theoretical foundations for inclusive digital rural policies and is vital for safeguarding smallholders' epistemic sovereignty.

Key words: smart agriculture, data elements, data poverty, algorithmic bias, socio-technical systems, digital inclusion, smallholder empowerment