The Use of Causal AI in State and Local Government Decision-Making

The views expressed are those of the author and do not necessarily reflect the views of ASPA as an organization.

By Bill Brantley
October 17, 2025

State and local governments are using data analytics and AI to inform decisions on budgeting and public programs. Traditional analytics often highlight correlations but these do not always indicate causation; relying on them can lead to ineffective or wasteful policies.

Causal AI represents an innovative methodology that seeks to advance beyond mere predictive analytics by elucidating the underlying reasons for observed outcomes. Through facilitating “what-if” scenarios and identifying genuine causal relationships, Causal AI enables agencies to make decisions that are both more effective and efficient.

What Is Causal AI (and How Is It Different from Traditional Analytics)?

Causal AI identifies cause-and-effect relationships, unlike traditional analytics which only find correlations. For instance, predictive models might link umbrellas to rain but can’t tell what causes rain. Because they depend on correlation, these models cannot predict the impact of changing a specific policy or distinguish causation from coincidence.

Causal AI combines artificial intelligence with causal inference methods from scientific research and program evaluation to address “what-if” and “why” questions. It develops models that illustrate how different factors interact and employs tools such as causal graphs and counterfactuals to examine the impact of changes. In contrast to predictive AI, Causal AI aims to explain outcomes by isolating the effects of individual factors, which can provide clearer and more dependable insights.

Making Government More Effective with Cause-and-Effect Insights

Shifting from correlation to causation helps governments design accountable, precise policies. Causal AI provides agencies with concrete insights into the real effects of programs, leading to better resource allocation and expansion of what works.
Governments can use Causal AI to test policies on historical data before implementation, helping officials predict outcomes and prevent issues. Using causal evidence leads to smarter decisions and improvements in safety, health, education and efficiency.

Case Study: Improving Education Outcomes with Causal AI

Consider a city school district testing a new tutoring program to boost math scores in struggling middle schools. Typically, results are measured by comparing test scores before and after tutoring or with schools not participating. While such analysis shows correlations, like a 5% rise in scores, it doesn’t prove tutoring caused the improvement. Other factors such as smaller classes or curriculum changes may be responsible. Standard analytics can’t confirm causation, so reports often attribute gains to the program with significant caveats.

Causal AI allows the district’s data science team to build a model of student performance by considering factors like test scores, attendance, tutoring hours and socio-economic indicators. They use causal inference to simulate what tutored students’ results would have been without intervention, essentially creating a virtual control group. By analyzing historical data and matching similar students, they find that tutoring leads to math scores rising by 5 percentage points compared to no tutoring. Causal AI thus accurately identifies the program’s true impact.

Comparing Traditional vs. Causal Approaches

In the past, districts observed score increases and considered possible causes, making decisions about tutoring with some uncertainty. Causal AI identifies the relationship between the tutoring program and measurable improvement, indicating which students or schools benefit most. This enables administrators to make decisions based on evidence, such as adjusting or expanding tutoring, rather than relying on assumptions. Causal analysis separates the program’s impact from other contributing factors, aiding policy choices that can affect costs and student achievement.

Getting Started with Causal AI: Practical Steps for Agencies

State and local administrators can start using Causal AI without being statistics experts but a careful approach is essential. Here are some practical steps to begin:

• Develop analytical skills: Provide training to equip analysts and decision-makers with causal inference methods. Even a basic workshop on “causation vs correlation” can improve your team’s ability to assess policies with evidence.
• Begin with a pilot project: Choose a policy or program in your agency that needs cause-and-effect analysis, such as a health campaign, policing strategy or education initiative. Make sure you have relevant outcome data. Starting small helps show value and gain support.
• Use existing tools and expertise: Instead of building a causal model from scratch, try open-source libraries like Microsoft’s DoWhy for step-by-step analysis. User-friendly Causal AI software is also available. Have data scientists explore these options or partner with universities or consultants if needed.
• Use interdisciplinary expertise: Involve data specialists and program experts to build credible causal models and interpret results. Collaboration ensures relevant factor selection and realistic causal assumptions.
• Ensure transparency and validate results: State assumptions, document your process and seek independent review for credibility. Establish standards for validation and transparency so analyses can be reviewed and improved. Start with simple, clear models before moving to more complex ones.

Causal AI helps public administration by identifying the causes of trends, enabling governments to test policies, target interventions and avoid false correlations. This leads to clearer, more responsible decisions. While new skills are needed to implement Causal AI, it allows for smarter resource allocation and more effective programs. Agencies using causal analysis will be better equipped to shape future policies.

---

Author: Dr. Bill Brantley is the President and Chief Learning Officer for BAS2A, an instructional design consultancy for state and local governments. He also teaches at the University of Louisville and the University of Maryland. His opinions are his own and do not reflect those of his employers. You can reach him at https://www.linkedin.com/in/billbrantley/
.

(1 votes, average: 5.00 out of 5)
Loading...