Kings have advisors, presidents have cabinets, CEOs have boards and TV show hosts have writers – every public figure relies on a cadre of trusted advisors for making decisions. Whenever crucial decisions are made, an army of astute specialists have spent countless hours researching, studying and preparing to communicate the most essential information to inform a decision maker on that issue. Without them, leaders would lead by instinct and most likely often get it wrong. What if these advisors were not human only but also AI-enabled decision systems?
This is what Modeling Religion Project is doing. Developed by a group of scientists, philosophers, and religion scholars, the project consists of a computer simulation populated by “virtual agents” mimicking the characteristics and beliefs of a country’s population. The model is then fed evidence-based social science tendencies of human behavior under certain conditions. For example, a sudden influx of foreigners may increase the probability of hostility by native groups.
Using this initial state as a baseline, they experiment using different scenarios to evaluate the effects of changes in the environment. Levers for change include adding newcomers, investing in education, changing economic policy among other factors. The model then simulates outcomes from the changes allowing for scholars and policy makers to understand the effects of decisions or trends in a nation. While the work focuses on religion, its findings have broad implications for other social sciences such as Psychology, Sociology and Political Science. Among others, one of their primary goal is to better understand what factors can impact the level of religious violence. The government of Norway is about to put the models to test, where they hope to use the insights of the model to better integrate refugees to their nation.
Certainly, a project of such ambition is not without difficulties. For one, there are ethical questions around who gets to decide what is a good outcome and what is not. For example, one of the models provides recommendation on how to speed up secularization in a nation. Is secularization a good path for every nation? Clearly, while the model raises interesting insights, using them in the real world may prove much harder than the complex math involved in building them. Furthermore, irresponsible use can quickly lead to social engineering.
While hesitation is welcome, the demand for effective decision making will only increase. Leaders from household to national levels face increasing complex scenarios. Consider the dilemma that parents face when planing for their children’s education knowing that future job market will be different from today. Consider organizational leaders working on 5-10 year plans when markets can change in minutes, demand can change in days and societies in the course of a few years. Hence, the need for AI-generated insights will only increase with time.
What are to make of AI-enabled advice for public policy? First, it is important to note that this already is a reality in large multi-national corporations. In recent years, companies have developed intelligent systems that seek to extract insights from the reams of customer data available to these organizations. These systems may not rise to the sophistication of the project above, but soon they will. Harnessing the power of data can provide an invaluable perspective to the decision making process. As complexity increases, intelligent systems can distill large amounts of data into digestible information that can make the difference between becoming a market leader or descending into irrelevancy. This dilemma will be true for governments as well. Missing data insights can be the difference between staying in power or losing the next election.
With this said, it is important to highlight that AI-enabled simulations will never be a replacement for wise decision making. The best models can only perform as well as the data they contain. They represent a picture of the past but are not suitable for anticipating black swan events. Moreover, leaders may have pick up signals of change that have not yet been detected by data collection systems. In this case, human intuition should not be discarded for computer calculations. There is also an issue of trust. Even if computer perform better than humans in making decisions, can humans trust it beyond their own capabilities? Would we trust an AI judge to sentence a person to death?
Here, as in other situations, it bears out to bring the contrast between automation and augmentation. Using AI systems to enhance human decision making is much better than using it to replace it altogether. While they will become increasingly necessary in public policy, they should never replace popular elected human decision-makers.
