Can predictive AI improve the efficacy of the G20?
The advent of predictive artificial intelligence opens new avenues for improving the effectiveness of the G20, offering fresh opportunities to drive progress on critical global issues and improve the well-being of people worldwide
Since its inception in 2008, the G20 summit has emerged as a pivotal international platform for addressing financial stability, economic development and global health challenges. Over the years, members have made thousands of commitments aimed at tackling these pressing issues. However, research reveals that only 54% of these commitments have been fully met, highlighting the need for innovative approaches to improve members’ compliance.
Traditional research has attributed compliance with G20 commitments to factors such as a member’s economic position and past track record of adherence to similar pledges. However, these underlying factors are largely immutable and offer limited leverage for enhancing compliance. The advent of predictive artificial intelligence opens new avenues for bolstering the G20’s effectiveness.
This article explores the potential of predictive AI in revolutionising the G20 decision-making process and driving higher rates of commitment fulfilment. Using a quantitative prediction model for compliance with G20 commitments with identified key summit and commitment characteristics that influence compliance probabilities, the G20 can strategically allocate resources and tailor engagement strategies for its members. With predictive AI as an ally, the G20 stands poised to create a more impactful, cooperative and successful platform for addressing global challenges and driving positive change on an international scale.
Predictive model
The predictive model used in this study is a random forest binary classifier, incorporating 500 trees and selecting eight variables at each split. To create this model, a diverse range of data sources was integrated, encompassing historic compliance rates, information on G20 ministerial meetings (including their timing and subjects), G20 members’ membership in international organisations, economic conditions, the subject of summit commitments, the host country of the summit and specific attributes of the commitments themselves – such as the strength of their language, mention of monetary commitments and specified timelines. The dataset used to train the model consisted of 7,940 assessments of past compliance patterns, collected by the G20 Research Group.
The model’s binary classification character allows it to predict whether a member is likely to comply with a given commitment. With a holdout dataset, the model exhibited an impressive 86% accuracy in its predictions for compliance. Attaining even higher accuracy could be challenging due to the influence of numerous idiosyncratic political and logistical factors that may be difficult to quantify in this context.
To facilitate ease of use and accessibility, the predictive AI tool has been developed as a user-friendly online web application. This application enables users to input specific features of a commitment made during a G20 summit and receive predictions on the likelihood of each member fulfilling that commitment. Users can also experiment with different settings, such as arranging relevant G20 ministerial meetings, to explore how these adjustments might increase the probability of members’ compliance.
The case for predictive modelling
Three features make a strong case for using this predictive model: it can identify high-risk commitments for G20 members early, it can provide real-time decision support during summits and it supports the ability to target resource allocations to those areas where compliance needs to increase the most.
Regarding early identification of high-risk commitments, predictive AI models can analyse the characteristics of commitments and detect patterns associated with low compliance probabilities. For instance, commitments that mention specific dates or monetary values have been shown to be less likely to be met. By recognising such characteristics early, the G20 can prioritise these commitments for targeted support and diplomatic efforts. Furthermore, the AI system can also identify members that historically demonstrate low compliance rates, allowing for tailored interventions and engagement strategies.
During G20 summits, leaders are often faced with complex decisions and limited time to address critical issues. Predictive AI can offer real-time decision support by continuously analysing the evolving dynamics of discussions and commitments made. This real-time analysis can empower leaders to make more informed and data-driven decisions, fostering a more productive summit outcome.
Resource allocation is a crucial aspect of enhancing compliance rates. By predicting the likelihood of commitment fulfilment for each member, the G20 can allocate resources strategically. Countries with a higher probability of non-compliance can receive targeted support to address potential obstacles to implementation. Conversely, members with a strong track record of compliance can be encouraged to share best practices and offer support to others who comply less well.
Conclusion
Integrating predictive AI into the G20’s operations has the potential to revolutionise the organisation’s approach to decision-making and commitment fulfillment. By harnessing the power of data and AI-driven insights, the G20 can identify high-risk commitments and members, make more informed decisions during summits, allocate resources strategically, foster cooperation and continuously improve its effectiveness. Ultimately, predictive AI can pave the way for a more effective and impactful G20, driving progress on critical global issues and improving the well-being of people worldwide.
The G20 Compliance Simulator is available at https://g20-utoronto.shinyapps.io/compliance-tool.
Full data and code are available at https://github.com/rapsoj/g20-compliance.
