What is agent-based modeling?

Over the last 10 years, the availability of more and more data, the development of low cost cloud computing, and the emergence of Machine Learning (ML) and Artificial Intelligence (AI) led to the creation of a new field: data science.

Data science allows organizations to automate decision-making related to questions like:

• Which ad should be served to a particular person based on their browsing history?
• Is this person likely to get heart disease in the next five years?
• Should I extend a mortgage to this person?

ML and other popular data science techniques answer these questions by identifying hidden trends in large data sets, but the success of these types of models depends on the future looking similar to the past.

Unfortunately, for many strategic questions, the future looks different from the past. We are living in an era of increasing uncertainty and organizations need to make decisions in an environment subject to:

• Technology change, e.g. the shift to electric cars or ChatGPT.
• Cultural change, e.g. the trend towards working from home.
• Economic change, e.g. the rise of inflation and interest rates.
• Demographic change, e.g. an aging population.
• Climate change, e.g. an increasing frequency of extreme weather events.
• Unprecedented shocks, e.g. COVID or the invasion of Ukraine.

So, how should organizations use data to make better strategic decisions where the future looks different from the past? For example, you might be trying to answer questions like:

• Which product categories and market segments will grow over the next five years?
• What might happen to the prevalence of heart disease if we reduce levels of sugar in food?
• How should we go-to-market in a new region?
• What sort of mortgage product will we need to support younger peoples’ needs?
• How should a governor respond to a new crisis?
• What does a particular shift in behavior mean for our business?
• How might changes in population health impact insurance premiums?

In this context, automating decisions by predicting the future is impossible, but improving decisions by exploring different possible futures is central to developing robust, effective strategies.

Agent-based modeling (ABM) is a data science technique that allows organizations to test answers to these “what if” questions at a granular level, evaluating alternative strategies over a range of scenarios to identify your best opportunity to create leverage. ABM complements predictive techniques, like ML, by enabling data scientists to provide actionable insights to improve decision-making around a new and expanding class of strategic questions.