The First Open-Source Equitable Decision Intelligence Model

When incidents are catastrophic and/or happen in compromised environments, complexity can increase rapidly and dramatically, compromising response objectives and resulting in catastrophic failure. The cost of these failures is measured in destruction and human lives, making even minimal reductions in capabilities untenable. A rapidly changing environment requires that the modern emergency manager is capable of quickly understanding community needs, including the needs of underserved populations and traditionally underrepresented groups.

Only if properly oriented to the conditions of four key environments could the modern emergency manager deliver equitable decisions before, during, and after disasters. This requires understanding pre-existing conditions of the natural, social, built, and economic environments. To do that, the following is needed: good data, good data processing, the ability to make sense of that data, and the ability to derive recommendations that are sensitized to the needs of the real people that live, work, and play in a community.

To disrupt inequity, four steps assist in the process: 

  • Step 1: Acknowledge the hidden bias¬†
  • Step 2: Understand and assess outcomes of bias
  • Step 3: Design equitable approaches
  • Step 4: Develop, test, and evaluate for equity practices

Current Decision-Making Process

Currently utilized technologies and protocols do not enable effective real-time, distributed, artificial intelligence (AI), machine learning, and other advanced analytics that encourage decision intelligence. Decision intelligence is the use of data from multiple sources to make quick and informed decisions that will support organizational effectiveness, response time, efficiency, and equity. To achieve this, a diverse pool of AI talent will be included to contribute to value sensitive design and training sets representative of social groups and their needs. Equity is intentional and must be prioritized early and often. AI should be beneficial to all of humanity, not harmful to the most marginalized. Diversity must be present in all phases of development to ensure proper perspective.

Disasters of all kinds affect daily life, so the programs focused on preparedness, mitigation, and outcomes for disaster survivors need to be equitable.

Currently, many decisions are being made based on frameworks, systems, policies, and programs that do not explicitly incorporate¬†equity into their process.¬†Information modeling, when applied to decision support, can provide recommendations on course of actions, in addition to playing out scenarios during the planning process.¬†Understanding equity and unintended biases in AI model-making processes is critical to developing an effective and inclusive model. For this collaborative effort,¬†equity is defined¬†by the Federal Emergency Management Agency as¬†‚Äúthe consistent and systematic fair, just and impartial treatment of all individuals.‚Ä̬†It is critical to¬†also acknowledge how bias, in some AI efforts has negatively impacted equity, such as seen across the public safety community, and how these lessons learned have directly informed an open-source model approach. Not being able to predict probabilistic scenarios for the event or outcome makes it difficult to design or implement effective interventions. Emergency managers need to mitigate disaster using predictive simulation to understand the outcome of an event.

It is important to understand:

  • Predictive modeling should be neutral.
  • Data often has bias.¬†
  • Algorithms must remove bias.
  • Recommendations should provide equitable benefit to the whole community.

In an era where disasters of all kinds are destructive to humans and their daily lives, the programs focused on preparedness, mitigation, and outcomes for disaster survivors are not always equitable, but that can change. It starts with understanding the makeup of communities and the nexus between social and economic vulnerability, social capital, and the impacts of emergency management decisions. This collective effort seeks to reverse industry trends and disrupt the impact of this human-made disaster that many people face and that has proven to exacerbate the impacts of hurricanes, floods, fires, and pandemics. 

Developing a Better Whole Community Model

This equitable model-building process should allow the authors to include in the model an understanding of the actual needs from the whole community, which can lead to a deeper understanding of the unique and diverse needs of a population ‚Äď including its demographics, values, norms, community structures, networks, and relationships. The model should include diverse community members, social and community service groups and institutions, faith-based organizations and disability groups, academia, professional associations, and the private and nonprofit sectors, while including government agencies who may not traditionally have been directly involved in emergency management.

During the modeling process, there needs to be an understanding of how human-centric decision intelligence can provide recommendations and courses of actions that assist the decision maker in understanding the current disaster environment. The proper calculation prepares communities for what comes next. The authors, as a group, have come together to design, develop, and implement the first open-source equitable decision intelligence model. The goal of this model is to be sensitive to social and economic vulnerabilities and supportive of advanced planning, preparedness, mitigation, response, and recovery needs. Model building is a team sport, requiring input from a diverse group of stakeholders to truly be effective. Models also need to be local, as are all disasters.

Technology is coded by human beings. If the humans developing the data and technology are introducing bias on the front end, then the back-end recommendations will be biased as well. Many groups that develop AI tend to do it in isolation, claiming they can solve the world’s problems while delivering incomplete and unsatisfactory or harmful results. For this reason, the authors have chosen to convene an effort that is fully open, inclusive, and crowd-tested, to develop something that is real, that is lasting, and that earns the trust and confidence of the whole community. This will ensure that both data equity and societal equity considerations are identified and accounted for on the front end of the design and development process, so that the back-end insights and recommendations will serve the whole community. The end goal is to publish an operational equity standard from which future decision making can be based in the context of preparedness, mitigation, response, and recovery.

The authors of this article are actively working to create and deploy explainable and equitable AI, enabling emergency management to assist in all aspects of decision support. Please contact with any questions or join the LinkedIn Group at To sign up for a pilot in a local jurisdiction, please access this form

Eric Kant

Eric Kant, Technical Translator, SPIN Global, LLC, Washington, DC.

Joel Thomas

Joel Thomas, Founder & CEO, SPIN Global, LLC, Washington, DC.

Chauncia Willis

Chauncia Willis, Co-founder & CEO, Institute for Diversity and Inclusion in Emergency Management, Atlanta, GA.

Sarah K. Miller

Sarah K. Miller, Region 10 President, International Association of Emergency Managers, Falls Church, VA.

Nissim Titan

Nissim Titan, CEO, 4Cast Inc., Missoula, MT, and 4Cast LTD, Petah Tikva, Israel.

Tzofit Chen

Tzofit Chen, Director of Marketing & Business Development, 4Cast Inc., Missoula, MT, and 4Cast LTD, Petah Tikva, Israel.

Brian Kruzan

Brian Kruzan: Senior Analyst. SPIN Global LLC, Washington, D.C.

Camila Tapias

Camila Tapias: Global Disaster Resilience Specialist: SPIN Global, LLC, Washington, DC.

Alexa Squirini

Alexa Squirini: Analyst, SPIN Global, LLC, Washington, DC.



No tags to display


Translate ¬Ľ