In a paper published in the journal Agriculture, researchers examined the role of artificial intelligence (AI) in food security, specifically focusing on stakeholder involvement in AI modeling for food security. They found that AI models were extensively used to assess various food security indicators across six continents, with a notable concentration of studies in sub-Saharan Africa.
Study: Leveraging Artificial Intelligence for Food Security: A Comprehensive Review. Image credit: Generated using DALL.E.3
AI research organizations, primarily based in Europe or the Americas, identified collaborations with local organizations and external partners. The analysis revealed three prevalent patterns in applying AI models for food security research: exclusive utilization for food security assessments, partial stakeholder involvement in the AI modeling, and stakeholder engagement through an iterative approach. The study concluded that research involving AI should consider incorporating stakeholder feedback and implementing research outcomes for addressing real-world food security challenges.
AI for Enhanced Food Security
AI technology is critical in addressing the intricate challenges of food security, driven by factors such as population growth, resource depletion, climate change, and increased food consumption. With smallholder farmers producing about 80% of food in developing countries, the application of AI, particularly in modeling, offers enhanced efficiency and automation. Stakeholder collaboration is essential for harnessing AI's potential, but challenges remain in translating model outputs into actionable policies and integrating local knowledge effectively. It underscores the need for stakeholder involvement and feedback mechanisms in AI modeling for food security.
AI-Based Modeling in Food Security: An analysis of the selected documents revealed the application of AI models for food security research in 68 countries across six continents. The distribution of the studies conducted in each continent is as follows: Africa (58 countries), Europe (17 countries), South America (16 countries), Asia (38 countries), North America (nine countries), and Australia (four countries). Additionally, some studies (30) had a global focus or focused on countries on different continents.
Among the 68 countries identified, approximately one-third of the studies were from the Global North, while two-thirds were from the Global South, with a particular emphasis on sub-Saharan African countries. Experimenters researched AI and food security in Europe, including Hungary, Spain, the UK, Italy, Poland, the Netherlands, Germany, and Greece.
AI-Based Food Security Indicators: Research on AI and food security has analyzed indicators, including accessibility, availability, affordability, and utilization. However, most studies have primarily focused on the availability aspect of food security. Regarding the accessibility indicator, 35 documents applied AI models to assess household food distribution disparities. Of the 171 documents analyzed, 105 mainly focused on food production, representing the availability indicator of food security. These studies explored various aspects of food availability, including the impact of climate change on crop production, soil health, pest and disease patterns, and agricultural productivity. Researchers also studied climate change and assessed the contribution of the livestock, aquaculture, and fisheries sectors to food availability. Moreover, studies focused on water availability, agricultural production systems, and farmers' behavior under the availability indicator.
The affordability indicator of food security in AI model studies was limited to 22 documents, which explored the effect of subsidy policies on the food security of rural households. AI models were also applied to assess food utilization, particularly regarding the consumption patterns of specific food products and their impact on hunger and rural health.
In this study, food security indicators were assessed at multiple levels, ranging from local to global.
Model Types: Various models were identified, with some studies applying a combination of models, including systemic and dynamic modeling. In analyzing natural systems, researchers also utilized machine learning algorithms and employed agent-based modeling approaches. Some models included AI algorithms at certain stages of the research.
Institutions Involved in AI Food Security Modeling: This study identified three categories of institutions engaged in AI modeling research for food security. These categories are local organizations, collaborations between local and foreign research institutions, and foreign organizations, including universities and international research institutions.
Foreign organizations primarily led AI modeling research on food security, often collaborating with universities and research institutions from the Global North and local communities in the Global South. Local organizations engaged in AI modeling research for food security were predominantly from developed countries, although there were exceptions.
Funding for AI Food Security Modeling: Researchers identified three funding categories for AI modeling research on food security: foreign, collaborative, and local funding. Development agencies, research institutions, state organizations, financial institutions, and other forums provided foreign funding. Joint funding involved contributions between local universities, research institutions, and foreign partners. Home-based institutions received local funding from governments through research councils.
AI Model Approaches in Food Security: Researchers adopted three main approaches in AI model research for food security. The first approach used AI models to analyze data without involving stakeholders, focusing on AI's potential and the models' limitations. The second approach involved integrating AI models, biophysical data, and primary data derived from stakeholders, strongly emphasizing building a modeling framework and understanding food security indicators. The third approach, applied in limited instances, involved stakeholders in an iterative process of AI modeling for food security, with researchers providing feedback to study communities and assisting in implementing the results. Stakeholder engagement, capacity-building, and monitoring and evaluation were critical aspects of this approach. Findings from these approaches offered insights into the application of AI in food security research and the importance of involving stakeholders in the process.
Conclusion
In conclusion, the analysis underscores the global reach of AI modeling research in food security and the importance of considering all aspects of food security, not just availability. Stakeholder engagement and community participation are crucial to enhancing trust and empowering marginalized groups. Recommendations include building trust, promoting open communication, and developing Decision Support System (DSS) tools. Future research should focus on measuring AI's effectiveness, adaptability, and ethical implications. These insights will contribute to more sustainable and equitable food security strategies.
Scaling Large Language Models Makes Them Less Reliable, Producing Confident but Incorrect Answers