Send message to Authors

Paper short abstract:

We conducted a landscape analysis of Artificial Intelligence in agriculture focusing on smallholder farming contexts in low- and middle-income countries. Findings suggest that academic research on AI effectiveness lags behind technological developments and practitioner studies.

Paper long abstract:

Introduction:

Use of Artificial Intelligence in smallholder agriculture has exponentially grown, but the effectiveness and implications for productivity, incomes, or livelihoods is poorly understood. We assessed the technical effectiveness of AI innovations and the socio-economic, ethical, and governance dimensions critical for AI implementation and scaling.

Methods:

- Rapid Review: Follow Cochrane Collaboration and Campbell Collaboration recommended methods. Qualitative evidence was critically appraised using the NICE framework; quantitative studies underwent evaluation using PROBAST, Cochrane ROB2, and ROBINS-I tools.

- Narrative Review: Capture emerging trends and implementation realities from developer and implementer perspectives, sources like blogs and editorials reviewed.

- Regional Case Studies: Reflect diverse applications of AI in agriculture with concrete regional examples of AI implementation and impact on local communities.

- Deep Dives: Examine socio-political and governance aspects of AI implementation, focusing on ethics: e.g. digital divide, digital inclusion.

- Stakeholder Consultations: Validate findings from reviews and case studies with agricultural sector and AI experts and stakeholders. Generate practical insights regarding the real-world application of AI innovations and implementation challenges.

Results: The rapid review synthesized 66 peer-reviewed studies: 42 quantitative, 20 qualitative, and 2 employing mixed methods. The narrative review includes 26 qualitative studies. Findings suggest (1) a mismatch between how AI is studied academically versus real-life implementation; (2) limited evidence for the effectiveness of AI in smallholder agriculture; (3) lacking attention for agricultural contexts in existing national and regional policies and regulations; (4) risks for maintaining or exacerbating the social, ethical, and sustainability challenges previously documented for digital agriculture.