AI, Neuroscience, and BI

Tauhid Nur Azhar

Photo by Alina Grubnyak on Unsplash

Being close friends with academic colleagues at the Bank Indonesia Institute, such as Kang Farid, the late Kang Oman, and Kang Janu Dewandaru (Dean of ALGM), has made me addicted to exploring economic issues, especially those related to the role of central banks. Bank Indonesia, as the central bank of the Republic of Indonesia, has functions including:

• Maintaining the stability of the rupiah value
• Maintaining the stability of the payment system
• Maintaining the stability of the financial system
• Managing crises
• Determining and setting payment instruments
• Providing approval for payment system operations
• Supervising payment system operations

In addition, Bank Indonesia also provides central banking services, such as: Giro accounts, L/C transactions, and Sub-Registry.

Apart from that, as a producer of regulations and policy maker in the monetary sector, such as setting the BI rate, etc., there are many economic issues that need to be considered by Bank Indonesia.

Many indicators need to be taken into account when formulating national economic policies. The dynamics of bilateral relationships in the region, up to global conflicts that directly and indirectly correlate with the supply chain of raw materials, energy sources, and commodity export market conditions, combined with the dynamics generated by natural or environmental factors such as global climate change, natural disasters, and political stability in a region, which in turn affect purchasing power, market absorption, and exchange rate changes, etc.

Related data, such as economic growth, micro and macroeconomic conditions, can be analyzed with the help of AI and produce targeted economic policies and regulations.

Micro and macroeconomic conditions in the fiscal, monetary, and real sectors are one of the most interesting topics to analyze.

Why? Because they are dynamic, volatile, unpredictable, and often explosive, causing devastating impacts. Some of us still remember the 1997–98 monetary crisis, don’t we?

Where the exchange rate plummeted out of control, and various fundamental indicators changed in an anomalous pattern. The random pattern triggered national, regional, and global panic.

The exponential change in exchange rates led to a radical depreciation of asset values and an irrational increase in import prices. Entrepreneurs in various sectors that rely on imported raw materials and traders and distributors of imported products almost simultaneously went bankrupt.

The microeconomy was severely shaken. Purchasing power parity plummeted, and the escalation of various economic pressures led to social problems. The explosion of various scales colored the economic turmoil, which in turn affected national stability and resilience.

Microeconomics itself is a branch of economics that studies consumer behavior, firms, and direct interactions. Microeconomic theory, sometimes simplified as price theory, analyzes how decisions and behavior affect supply and demand.

Microeconomics focuses more on analyzing how to allocate resources to achieve the right balance.

The scope of microeconomics includes: interactions in goods markets, seller and consumer behavior, quantity, quality, and price of goods or services in the market.

Some examples of microeconomic applications are: demand and supply of goods, individual vs. aggregate income, employment and unemployment, inflation rate, and investment decisions.

Meanwhile, the study of policy and regulation implementation in the fiscal and monetary sectors, aimed at maintaining national stability, falls under the study and policy of macroeconomics.

Macroeconomics is a branch of economics that studies the economy as a whole, including markets, businesses, consumers, and governments.

Macroeconomics examines broader economic phenomena, such as unemployment rates, national income, economic growth rates, inflation, and price levels.

Macroeconomics differs from microeconomics, which focuses on analyzing how to allocate resources to achieve the right balance.

Macroeconomics focuses more on analyzing the impact of economic activities on the overall economy.

The goal of macroeconomic development is to increase and maximize economic growth, which aims to achieve equality and stability.

Currently, to maintain economic growth and control inflation, for example, the central bank has a team of experts, including economists like Josua Pardede, who conduct research and provide input to the BI Governor’s Council.

The results of their research are used to formulate policies and strategies, which are then communicated to the public through the BI Governor’s Council meeting.

The data presented to the public as a result of the BI Governor’s Council meeting, for example, includes information such as:

• BI-Rate set at 6.25%
• Deposit Facility interest rate set at 5.50%
• Lending Facility interest rate set at 7.00%

This decision is consistent with the pro-stability monetary policy, which is a preemptive and forward-looking measure to ensure that inflation remains under control within the target range of 2.5±1% in 2024 and 2025, including the effectiveness of maintaining foreign capital inflows and Rupiah exchange rate stability.

Meanwhile, macroprudential policy and payment systems remain pro-growth to support sustainable economic growth. The macroprudential policy is relaxed to encourage bank credit and financing to the business and household sectors.

Payment system policy is directed towards strengthening the reliability of infrastructure and industry structure, as well as expanding digitalization of payment systems.

To ensure stability and support sustainable economic growth amidst high uncertainty in the global financial market, Bank Indonesia continues to strengthen its monetary, macroprudential, and payment system policies through:

• Optimizing the use of Rupiah Securities (SRBI), Foreign Exchange Securities (SVBI), and Sukuk Valas Bank Indonesia (SUVBI).
• Stabilizing the Rupiah exchange rate through interventions in the foreign exchange market, Domestic Non-Deliverable Forward (DNDF), and Government Securities (SBN) in the secondary market.
• Strengthening the term-repo SBN and competitive swap valas strategy to maintain banking liquidity, accompanied by deepening transparency policies on the Basic Credit Interest Rate (SBDK) with a focus on sectoral economic development.
• Strengthening synergy with payment system industry players to expand QRIS merchant acquisition in all UMKM categories through improved service quality, promotional programs, and QRIS usage campaigns, including QRIS Jelajah Indonesia.

Indonesia’s Balance of Payments (BOP) remains good, supporting external resilience. The current account deficit in the first quarter of 2024 remains low, supported by the continued surplus in the trade balance of goods.

Meanwhile, the capital and financial account in the first quarter of 2024 recorded a deficit, in line with the uncertainty in the global financial market.

The latest developments in the second quarter of 2024 show that the BOP has improved, supported by the continued surplus in the trade balance in April 2024 of $3.6 billion, driven by non-oil and gas exports.

Meanwhile, the net inflow of portfolio investment in the second quarter of 2024 (up to May 20, 2024) was $1.8 billion, driven by the positive impact of the monetary policy mix response of Bank Indonesia.

Indonesia’s foreign exchange reserves at the end of April 2024 remained high at $136.2 billion, equivalent to 6.1 months of imports or 6.0 months of imports and government foreign debt payments, and above the international standard of 3 months of imports.

Analysis of various variables from the above data can be done more optimally by adding enrichment factors based on neurophysiological and neurobiological functions of human behavior as the main actor in every economic event.

Where analysis based on neuroscientific co-factors can serve as a reference for the development of a Smart Central Bank System using AI technology.

As input for the AI system that can be developed by BI as the central bank of Indonesia, it would be beneficial to consider factors from behavioral economic theories, such as those developed by Kahneman and Tversky, as part of the assessment of group and individual behavior that contributes to various economic interaction models.

Prospect Theory is one of the main contributions of Daniel Kahneman and Amos Tversky in the field of behavioral economics.

This theory explains how people make decisions in situations of risk and uncertainty, challenging the assumptions of expected utility theory in classical economics.

The following are some of the main components of Prospect Theory:

1. Value Function, where people tend to evaluate outcomes based on changes from a certain reference point, rather than based on their final wealth. The value function is concave for gains (representing diminishing marginal utility) and convex for losses (representing increasing marginal disutility). This function is steeper for losses than for gains, indicating that people experience loss aversion.
2. Probability Weighting, where people tend to overestimate the probability of low-probability events and underestimate the probability of high-probability events. Probability weighting is non-linear and shows bias in risk assessment.

Some principles of Prospect Theory include:

• Endowment Effect, where people tend to give higher value to something they own compared to if they didn’t own it, showing aversion to loss.
• Certainty Effect, where people tend to give disproportionate weight to certain outcomes compared to uncertain outcomes, even if the expected utility is the same.
• Isolation Effect, where people tend to focus on the differences between available options rather than their similarities, often leading to inconsistent decision-making.

All economic theories and analysis models cannot be separated from the role of neurophysiology in decision-making. Therefore, understanding various neuroscientific functions is a key aspect that needs to be understood in regulatory planning and policy development, including the development of AI-based decision-making systems.

Research in neuroscience has helped explain the brain mechanisms underlying Prospect Theory and other economic behaviors. Some of the brain functions and areas involved include:

• Dorsolateral Prefrontal Cortex (DLPFC), which is related to decision-making and behavioral regulation, including self-control and risk evaluation.
• Amygdala, which plays a role in emotional processing, particularly related to fear and loss aversion.
• Ventromedial Prefrontal Cortex (vmPFC), which is related to value evaluation and emotional integration into decision-making.
• Striatum, which is involved in reward processing and probability evaluation, influencing how individuals respond to potential gains and losses.
• Insula, which is related to intuitive and emotional feelings, as well as reactions to uncertainty and risk.

Neuroscience supports the idea that economic decisions are not entirely rational and are influenced by emotional and psychological factors. These findings are consistent with Prospect Theory, which shows that:

• People are more likely to avoid losses than to pursue gains.
• Risk assessment is non-linear and influenced by cognitive biases.

The use of AI in economic analysis at central banks can provide deeper insights and more accurate predictions for better decision-making. By combining various AI models such as NLP, machine learning, deep learning, and agent-based modeling, central banks can more effectively control exchange rates and inflation, ultimately supporting economic stability.

Integrating AI into this process not only increases prediction accuracy but also provides flexibility in facing global economic uncertainty.

AI, specifically machine learning, can be used to predict inflation and exchange rates. Models that can be used include Long Short-Term Memory (LSTM) and Recurrent Neural Networks (RNN).

LSTM and RNN are very effective in handling time series data, such as inflation and exchange rate data. These models can learn patterns in historical data and predict future values.

More accurate predictions about inflation and exchange rates can help central banks plan timely and effective monetary policies.

The use of AI in central banks can also be used to develop market sentiment analysis systems, including using Convolutional Neural Networks (CNN), Transformer models (BERT or GPT), which can analyze sentiment from news, social media, and financial publications. By understanding market sentiment, central banks can evaluate the impact of policies that have been or will be implemented.

The model can provide insights into how the market responds to monetary policy and predict potential risks or benefits of the policy.

AI can be used to detect economic anomalies, such as unusual spikes in inflation or exchange rate fluctuations, through the application of Autoencoders, Variational Autoencoders (VAE).

The autoencoder model can detect anomalies in economic data, allowing central banks to take quick action to address potential crises through early detection of economic problems that can affect monetary stability.

Monetary policy simulations can be performed using Generative Adversarial Networks (GANs) or Reinforcement Learning (RL). GANs can be used to simulate different economic scenarios based on historical data and future predictions. Meanwhile, reinforcement learning can be used to evaluate optimal monetary policies through simulation.

This model can provide a picture of the impact of various monetary policies before they are implemented in the real world, allowing for adjustments and optimization of policies.

Complex economic modeling can use Graph Neural Networks (GNN) and Deep Reinforcement Learning.

GNN can be used to model complex interactions between various economic agents (banks, companies, governments, individuals) and understand the impact of policies at the micro and macro levels.

Meanwhile, deep reinforcement learning can help find optimal monetary policies through repeated simulations.

The benefits of using Deep Learning and its applications in monetary policy planning include:

• More Accurate Predictions: By predicting inflation and exchange rates more accurately, central banks can adjust interest rates and other monetary policies to control inflation and exchange rate stability.
• Rapid Response to Market: Sentiment analysis helps understand market reactions to proposed or implemented policies, allowing for real-time policy adjustments.
• Risk Reduction: Anomaly detection enables early identification of economic problems that can be intervened before they develop into crises.
• Policy Optimization: Simulation and complex economic modeling enable central banks to try out various policy scenarios and choose the best one based on simulation results.

Some supporting economic theories that can be used in the development of AI in central banks include:

By leveraging the capabilities of deep learning to process large and complex data, central banks can make more effective and targeted decisions, ultimately supporting economic stability.

The stages of developing an AI-based smart system for a central bank can be described as follows:

Data Collection and Preprocessing

• Collecting data from various sources (economic, financial, and social) and cleaning it to ensure data quality.
• Technologies used include Big Data Analytics and ETL (Extract, Transform, Load) tools.

Model Development and Training

• Developing and training AI models using historical data and relevant economic indicators.
• Technologies that can be used include TensorFlow, PyTorch, and Scikit-learn.

Model Evaluation and Validation

• Evaluating model performance using metrics such as RMSE (Root Mean Square Error) for predictive models and AUC (Area Under Curve) for classification models.
• Technologies that can be used include Cross-validation techniques and hyperparameter tuning.

Deployment and Monitoring

• Implementing the model in the central bank’s operational system and continuously monitoring its performance.
• Technologies that can help include Cloud computing (AWS, Azure) and monitoring tools (Prometheus, Grafana).

There is great hope that the integration of fundamental neuroscience knowledge about decision-making systems underlying economic behavior theory, with AI-based smart analysis technology, will be able to improve the performance of the central bank.

Regulatory planning and monetary and macroeconomic policy will be more targeted, effective, optimal in resource utilization, and can be done quickly. Considering the crucial aspect of anticipating global and local economic dynamics is the speed and accuracy of decision-making, accompanied by powerful data collection and processing capabilities.