Thought Leadership | Blog Posts

AI Crisis Response vs Human Judgment

Written by Seth Mattison | Jun 7, 2026 2:27:26 AM

When crises strike, decisions must be made fast. AI can process vast data in seconds, while humans bring ethics, experience, and emotional intelligence. The best approach combines both: AI handles data-heavy tasks, and humans make judgment calls. This partnership ensures speed, accuracy, and accountability.

Key points:

  • AI excels at analyzing data, spotting patterns, and automating repetitive tasks (e.g., wildfire detection or supply chain disruptions).
  • Humans are essential for ethical decisions, public trust, and adapting to new, unpredictable crises.
  • Challenges exist: AI can falter with biases or infrastructure failures, while humans face stress, cognitive overload, and biases like confirmation or status quo.

Effective crisis response requires:

  1. Clear roles: Use human-in-the-loop for high-risk tasks; human-on-the-loop for faster, lower-risk tasks.
  2. Governance: Oversight, bias checks, and escalation protocols ensure safe AI use.
  3. Training: Leaders need skills in data literacy and ethical decision-making to work seamlessly with AI.

AI in Crisis Response: Strengths and Limitations

Where AI Adds Value in Crisis Management

When disasters strike, chaos often follows, with overwhelming amounts of data pouring in from sources like satellite imagery, social media, emergency calls, and sensor alerts. Human teams can only process so much at once. This is where AI steps in, transforming raw data into actionable insights in seconds. Researchers describe this as "clarity in motion", giving decision-makers real-time clarity in high-pressure situations [1].

The numbers tell the story. AI-driven systems have been shown to enable 40% faster recovery times, speed up responses to supply chain disruptions by nearly 65%, and improve emergency medical response times by 10%–20% through smarter call triage and optimized ambulance routing [5]. Beyond speed, AI helps retain and analyze lessons from past crises, ensuring that valuable insights from After Action Reviews (AARs) aren't lost when leadership changes [1].

A great example of this is Cal Fire's deployment of AI in September 2023. Using image recognition across over 1,000 cameras, the system detected wildfires faster than human 911 callers, allowing for quicker containment and minimizing damage [6].

The Risks and Limits of AI in a Crisis

While AI offers impressive benefits, it isn't without its flaws. Its effectiveness depends heavily on the quality of its training data. When faced with scenarios outside its training, AI models can falter - a phenomenon known as overfitting. This was evident during the early days of the pandemic when response models struggled with the unexpected nature of the crisis [7].

Another critical issue is algorithmic bias. In October 2019, a study published in Science revealed that an AI algorithm managing health risks for 200 million Americans was racially biased. By using healthcare costs as a stand-in for health needs, the system unfairly disadvantaged Black patients, requiring them to be much sicker than White patients to receive equivalent care [7][8]. In emergencies, such biases can have amplified consequences.

AI's dependence on infrastructure like 4G/5G networks and substantial computing power also presents vulnerabilities. Power outages or network failures can render these systems useless precisely when they're most needed [10]. Additionally, automation bias - the tendency to trust machine-generated decisions without question - can create dangerous situations.

"The machine says 'Expectant.' It says it instantly, consistently, and with a confidence interval. None of those properties make it correct." - Chet Shermer, MD, Professor of Emergency Medicine [10]

These challenges highlight the importance of keeping human judgment and oversight at the center of AI-driven crisis management.

Governance and Design for Responsible AI Use

To address these risks, human oversight is essential. One of the biggest hurdles is the "many hands" problem, where accountability becomes murky when thousands of people interact with an AI system [3].

For responsible AI use in crises, certain principles are non-negotiable. First, systems should have adjustable autonomy, allowing human operators to reduce AI control when situations become unpredictable [3]. Second, continuous bias auditing must be a priority, with regular checks on how the system performs across different demographic groups - not just at launch, but throughout its use [7]. Finally, organizations that establish dedicated oversight roles, like algorithm auditors, are 60% more likely to achieve their project goals than those that don't [6].

"Successful implementation requires complementary investments in organizational capacity, data infrastructure, workforce training, community engagement, and continuous evaluation." - Elizabeth Campbell, PhD, Johns Hopkins Center for Outbreak Response Innovation [7]

The aim isn't to slow AI down but to ensure it operates reliably when it matters most. By stress-testing systems, clarifying accountability, and keeping humans in the loop, we can strike the right balance between efficiency and safety.

sbb-itb-9ceb23a

Human Judgment in Crisis Response: Strengths and Challenges

What Human Judgment Offers in a Crisis

AI excels at processing massive amounts of data, but when a situation veers into uncharted territory, it struggles to prioritize effectively. Human leaders, on the other hand, can adapt in real time - shifting focus from short-term objectives to critical, life-saving decisions. They use their ability to make sense of incomplete or conflicting information to impose order on chaos. For example, a leader can decide that saving lives takes precedence over protecting quarterly profits - something AI can't do because it's confined to its programmed parameters [11][3].

Another key strength is relational intelligence. Studies show that people trust crisis communications more when they come from a human rather than an AI system [2]. In moments of fear and uncertainty, people want to feel understood by another person - not a machine.

"You don't risk people's lives for efficiency. Only a human can make that call." - Oleksandr Balanutsa, Ukraine's Ambassador to the UAE [12]

Cognitive Biases and Emotional Pitfalls

While human judgment is invaluable, it isn’t flawless. In high-pressure situations, fast, instinctive decision-making - known as System I thinking - often takes over. This type of reasoning is quick but prone to errors. Stress and tight deadlines can crowd out slower, more deliberate thought processes, leading to mistakes [14].

The USS Vincennes tragedy in 1988 highlights this vulnerability. Under extreme pressure, the ship's captain misidentified Iran Air Flight 655 as a hostile aircraft, resulting in the loss of 290 civilian lives. This was a classic example of how stress can narrow focus and push speed over accuracy [14].

Cognitive overload makes things worse. For instance, critical care clinicians may face over 100 clinical decisions while managing just six patients. This sheer volume of decisions creates "noise" - random inconsistencies in judgment caused by fatigue, mood, or other unrelated factors [15][18]. The table below outlines common biases that often surface during crises:

Bias Type How It Shows Up in a Crisis
Confirmation Bias Favoring information that supports initial assumptions while ignoring contradictory evidence [16]
Status Quo Bias Clinging to the current plan even when new data suggests a better course of action [15]
Hindsight Bias Believing the crisis was more predictable than it actually was after events have unfolded [15]
Occam's Razor Fallacy Oversimplifying by defaulting to the easiest explanation and missing deeper complexities [15]

A growing concern is cognitive surrender - a tendency to accept AI-generated decisions without questioning them.

"Cognitive surrender is the moment when AI is not just doing a specialized task but making the decision, and the person adopts that decision as their own without recognizing the transfer has occurred." - Gideon Nave, Professor, The Wharton School [17]

To counter these challenges, organizations need to refine their decision-making processes and frameworks.

Building a Human Moat for Crisis Leadership

Addressing the issues of cognitive bias and emotional pitfalls, the Human Moat framework focuses on ethical reasoning and trust as the foundation for better decision-making. This concept, developed by Seth Mattison, emphasizes the qualities that set humans apart from machines: ethical reasoning, relational trust, effective decision-making under uncertainty, and a clear sense of purpose [sethmattison.com].

In practice, this means treating decision-making as a skill to be honed, not just an instinct. Organizations that invest in structured approaches - like scenario-based training, decision memos, and override protocols - are better equipped to handle real crises. For instance, BAE Systems demonstrated this in March 2026 by using case-based learning simulations to improve leadership agility and cross-departmental coordination in uncertain situations [4].

The aim isn’t to slow down decision-making but to ensure that quick decisions are grounded in sound judgment. Strengthening human judgment not only reduces the risk of bias but also complements AI’s analytical strengths, creating a more balanced and effective approach to crisis management.

Human vs AI: Who Decides During a Cyber Crisis?

AI vs. Human Judgment: A Side-by-Side Comparison

AI vs. Human Judgment in Crisis Response: Strengths, Limits & Key Stats

Knowing when to rely on AI and when to lean on human judgment is key to effective crisis management. These two forces aren't competitors but partners, each excelling in different areas. Choosing the right one at the right time can mean the difference between managing a crisis effectively and facing serious consequences. Here's a closer look at where each thrives.

Tasks Where AI Outperforms Human Judgment

AI stands out for its speed and ability to handle massive amounts of data. Take the January 2025 Eaton and Palisades wildfires in Los Angeles County as an example. AI-driven analysis identified coordination issues across four to eight different jurisdictions, delivering 38.93% higher accuracy in assessing damage severity from satellite and UAV images compared to human teams. It also maintained 60.94% greater consistency in decision-making across various scenarios [1] [9]. In contrast, human volunteers often experience "disaster fatigue", which can impair their decision-making over time.

For tasks that are repetitive, data-intensive, or time-sensitive - like monitoring sensors, drafting alerts, or highlighting anomalies - AI ensures there’s no delay between detecting a problem and responding to it [20].

"AI in critical event management is not about removing humans from the loop. It is about removing the friction between the signal and the structured response." - Dr. Shalen Sehgal, Crises Control [20]

However, while AI dominates structured, data-driven tasks, it struggles in situations that require creativity or ethical judgment.

Decisions That Require Human Leadership

Some situations demand the flexibility and ethical understanding that only humans can provide. AI falters in novel crises - those without historical data to guide its algorithms. This limitation can lead to "overfitting", where AI relies on outdated patterns that no longer apply [11]. Human leaders, on the other hand, can adjust strategies on the fly, interpreting public sentiment and making decisions that align with evolving circumstances.

Ethical dilemmas are another area where humans must take the lead. For instance, during a high-level panel at Rabdan Academy in May 2026, UAE officials explained how their missile defense system uses AI to monitor multiple threat signatures. However, the final decision to intercept a missile - and potentially risk civilian lives - remains with a human operator. This ensures moral reasoning is part of the process [12].

Comparing Strengths of AI and Human Judgment

The table below highlights where AI and human judgment excel:

Task Best Handled By Why
Real-time data monitoring AI Handles vast data streams without fatigue [1] [20]
Damage assessment from imagery AI Achieves 38.93% higher accuracy in structured tasks [9]
Drafting notifications and alerts AI Follows pre-approved policies instantly, reducing delays [20]
Ethical trade-offs (civilian safety) Human Requires moral judgment that AI lacks [12]
Unprecedented event response Human Adapts to new situations where no historical data exists [11]
Public communication and trust Human Builds credibility through empathy and accountability [11]
Pattern recognition in stable data Collaborative Human-AI teams achieve higher accuracy together [21]

The bottom line? AI provides the insights, but humans make the decisions. As Oleksandr Balanutsa, Ukraine's Ambassador to the UAE and former Deputy Defense Minister, aptly said:

"AI can be a player, a counselor, an advisor, but it cannot be the one making decisions." [12]

When lives are on the line, responsibility cannot be automated. The ultimate call - the one that carries accountability - belongs to humans. This balance between AI's capabilities and human judgment forms the backbone of a responsive and effective crisis management strategy.

How to Build Effective AI–Human Collaboration in Crisis Management

Building effective collaboration between AI and humans during crises requires systems that can handle pressure while maintaining clear roles and responsibilities.

Operating Models for AI–Human Integration

When integrating AI into crisis response workflows, two main models stand out. The choice between them depends on the level of risk involved.

In a human-in-the-loop model, every action requires human approval before execution. While this approach can slow things down, it provides a critical layer of oversight, making it ideal for high-stakes decisions like issuing apologies or admitting liability. On the other hand, a human-on-the-loop model allows AI to operate at machine speed, with humans stepping in only when specific thresholds are met. This is better suited for fast-moving, lower-risk tasks such as detecting bots or analyzing logs. However, this model demands robust safety mechanisms to ensure control.

Feature Human-in-the-Loop Human-on-the-Loop
Speed Slower; depends on human input Faster; operates at AI speed
Risk Level Lower; manual oversight at each step Higher; requires reliable override systems
Best Use Case High-risk decisions (e.g., liability admissions) High-speed tasks (e.g., bot detection)

An executable runbook serves as the practical tool that bridges these models. This dynamic guide outlines which tasks AI handles independently and which require human involvement, assigning real-time ownership and ensuring smooth coordination [23].

With operational models in place, the next critical step is establishing strong governance.

Setting Governance Rules and Guardrails

Effective governance ensures that human oversight remains central to decision-making. One proven method is the AI-Augmented Crisis Decision Matrix (ACDM), which organizes decision-making authority into three risk tiers:

  • Green (low risk): Pre-authorized staff handle decisions.
  • Yellow (moderate risk): Senior leads take charge.
  • Red (high risk): Decisions escalate to the C-suite and legal teams [22].

This framework allows teams to act quickly while minimizing unnecessary escalations.

To make oversight meaningful, it must have four key qualities [23]:

  • Legibility: Humans need to understand why the AI made a recommendation, not just what it suggested.
  • Accountability: A specific person must always own the final decision.
  • Interruptibility: The ability to pause or override AI actions at any point is essential.
  • Escalation: Uncertainty - not just errors - should trigger human review.

Without these safeguards, governance risks becoming a mere formality. For example, a 2023 study highlighted the danger of automation bias: when AI recommendations included high confidence scores, human reviewers often agreed with incorrect suggestions, even when they had the data to identify mistakes [23].

"You should design your AI so that uncertainty escalates to humans, not just errors." - Asya Bar-Ziv, Product Manager, Cutover [23]

The fallout from Vanderbilt University's February 2023 incident underscores the importance of thoughtful governance. After a mass shooting at Michigan State University, Vanderbilt's Peabody College issued a solidarity statement that revealed assistance from ChatGPT. The disclosure sparked backlash, with students calling the message "disgusting" and "insincere", turning a well-meaning gesture into a reputational crisis [2]. This case highlights that even appropriate AI use can backfire if not disclosed carefully.

Strong governance equips leaders to harness AI effectively without compromising sound judgment.

Preparing Leaders for AI-Augmented Decision-Making

The real challenge lies in preparing leaders to work seamlessly with AI. Beyond operational models and governance, leaders need to develop skills in data literacy, sound judgment, and adaptive thinking.

  • Data literacy allows leaders to critically evaluate AI outputs, such as understanding confidence levels, recognizing outdated patterns, and identifying when results may be flawed.
  • Sound judgment involves applying ethics, experience, and organizational context to decisions that AI alone cannot resolve.
  • Adaptive thinking shifts the focus from reactive crisis management to proactive planning. As Seth Mattison (sethmattison.com) explains, leaders should act as "pre-crisis governance architects", designing frameworks that empower teams to make decisive moves before crises escalate [22].

"Human–AI teams must train together before high-stakes decisions put the partnership to the test." - Dan Dworkis, M.D., Ph.D., Emergency Physician [13]

The urgency is clear: 57% of organizations currently lack decision-making maturity, rarely teaching leaders the skills or providing the tools needed for effective decision-making [4]. With AI projected to augment or automate half of all business decisions by 2027, organizations must act quickly to close this gap [4]. Those that invest in simulation-based training, updated decision rights, and leaders equipped to manage AI as a "silicon teammate" will be better prepared to navigate high-pressure situations successfully.

Conclusion: What Crisis Leadership Looks Like Going Forward

The message throughout this article is straightforward: neither AI nor human judgment can fully address the complexities of crisis leadership on their own. The most successful organizations combine AI's ability to work at speed and scale with the ethical and accountable decision-making that only humans can provide. As Oleksandr Balanutsa, Ukraine's Ambassador to the UAE, aptly put it:

"If the system recommends a wrong decision, the responsibility lies with the person who presses the button." [12]

To strike this balance, leaders must rely on distinctly human capabilities to guide AI-driven responses. While AI can draft reports, simulate scenarios, and highlight risks, the true advantage lies in a leader's ability to interpret situations, take responsibility, and offer strategic vision - something no algorithm can replicate. Seth Mattison (sethmattison.com) describes this as operating at the top of the value stack, where human skills like interpretive clarity and relational trust become the cornerstone of competitive strength.

"Human judgment is not a relic of a pre-digital era. It's the ultimate competitive advantage in an AI-driven world." [19]

The organizations that will thrive in future crises are those that integrate AI systems before a crisis occurs, train human–AI teams in low-pressure scenarios, and establish governance frameworks that ensure clear and accountable oversight. Decisions made during high-stakes crises are among the least likely to be automated, with an average automation risk of just 9.8% [24]. This underscores the enduring value of human oversight - not as a temporary safeguard, but as a critical, long-term strategic asset. The challenge isn't choosing between AI and human judgment; it's about investing in the structures, skills, and culture that allow both to excel when it matters most.

FAQs

When should AI make decisions vs. humans?

AI excels at handling tasks that demand speed, consistency, and the ability to recognize patterns - think routine workflows or data-heavy operations. On the other hand, humans bring irreplaceable value to areas like ethical decision-making, nuanced judgment, and managing high-pressure crises.

For decision-makers, it’s helpful to divide responsibilities into three categories: routine tasks are best handled by AI, unclear or complex situations benefit from a partnership between AI and humans, and new or critical challenges require human leadership, with AI playing a supporting role by offering analysis and insights.

How do you prevent bias and automation bias in crisis AI?

To minimize bias, think of AI as a helpful assistant rather than the ultimate authority. Always apply critical thinking to evaluate its outputs and make sure they align with your objectives. Divide tasks wisely: allow AI to manage routine, predictable decisions, while keeping more sensitive or high-impact scenarios under human supervision. Keep a record of instances where AI outputs are overridden, along with the reasoning behind those decisions, to improve future processes. Additionally, establish clear guidelines for when human intervention is required, particularly in situations involving ethical complexities.

What should leaders train for to work well with AI in a crisis?

To navigate crises effectively alongside AI, leaders need to cultivate distinctively human abilities - what some call a Human Moat - to balance and enhance AI's capabilities. These skills include:

  • Evaluative judgment: Assessing complex situations and determining what truly matters.
  • Tradeoff judgment: Weighing competing priorities and making balanced decisions.
  • Anticipatory judgment: Foreseeing potential outcomes and preparing for them.
  • Contextual judgment: Understanding the broader environment and adapting accordingly.
  • Editorial judgment: Refining and communicating information effectively.

In addition to honing these skills, leaders should focus on mastering decision triage, developing robust governance frameworks, and engaging in collaborative simulations. Emotional intelligence plays a critical role here, helping leaders uphold ethical standards, protect trust, and ensure accountability when the stakes are high.

Related Blog Posts