As National Intelligence College's Dr. Deborah Pfaff recently noted, the character of intelligence is changing and computational sciences are a contributing factor to such transformation. Traditional intelligence disciplines such as human intelligence (HUMINT), signals intelligence (SIGINT), geospatial intelligence (GEOINT), measurement and signature intelligence (MASINT), and open-source intelligence (OSINT) remain indispensable, yet they are increasingly challenged by adversaries that employ technological means to shape tactical operations and strategic environments below the threshold of armed conflict. Contemporary competition is becoming ever more focused upon discerning how adversaries perceive, decide, adapt, and influence others. Thus, as warfare progressively entails persistent competition in the cognitive domain, intelligence, surveillance and reconnaissance (ISR) must likewise evolve.

Several years ago, the concept of neurocognitive intelligence (NEURINT) was proposed as a complementary intelligence toolkit and set of practices for fortifying understanding and facilitating the neural, cognitive, emotional, and behavioral determinants that shape individual and collective decision-making. In our model, NEURINT comprises: 1) utilizing information from neuroscience, psychology, behavioral economics, sociology, anthropology, computational modeling, and intelligence analysis to characterize the cognitive terrain upon which tactical engagement and strategic competition increasingly occurs; and 2) employing techniques and technologies of neuroscience to facilitate and fortify the capabilities of ISR operators to discern, identify, define and analyze mission-valuable signal(s) within a background of informational noise. 

Dr. Pfaff’s acknowledgement that recent advances in generative agentic artificial intelligence, coupled with unprecedented access to large-scale heterogeneous datasets, now afford new means and methods that synergize both informational acquisitional and analytic aspects of NEURINT in ways that were previously aspirational. These technologies enable faster analysis and construction of adaptive computational ecosystems capable of modeling human cognition, forecasting behavioral trajectories, identifying vulnerabilities and opportunities, and supporting proactive strategic deterrence; and such capability may position NEURINT to be particularly well suited for left-of-bang deterrence (viz., aimed at disrupting, delaying, degrading, disincentivizing, and/or redirecting adversarial activities before hostile action occurs).

ISR Assessment in the Cognitive Era

Historically, the intelligence community has concentrated upon material indicators such as troop movements, weapons development, financial transactions, communications, infrastructure, and political activities. While these certainly remain critical, tactical maneuvering toward strategic competition increasingly depends upon variables that conventional intelligence often struggles to characterize. For example, adversaries may seek to influence beliefs, shape perception, manipulate narratives, alter public trust, exploit cognitive biases, amplify uncertainty, and induce behavioral change. Thus, intelligence must increasingly understand how adversaries think. 

NEURINT addresses this challenging task; its principal objective is to characterize the neurocognitive architecture underlying decision-making across multiple scales (e.g., from individuals and small leadership groups to larger collectives including organizations, populations, and societies). Such assessment encompasses a number of factors, as depicted in Table 1, below. These variables collectively constitute what may be termed the cognitive battlespace.

Table 1: 
NEURINT Foci of Assessment 

• Cognitive biases
• Emotional drivers
• Motivational factors and structures
• Cultural cognition
• Ideological factors and commitments
• Organizational dynamics
• Idiosyncratic and collective stress responses
• Idiosyncratic and systemic risk tolerance
• Patterns and profiles of collective moral reasoning
• Information-processing strategies
• Idiosyncratic and collective adaptive learning behaviors

Generative Agentic AI as a Force Multiplier

Recent developments in generative AI extend well beyond conversational language models. The emerging paradigm involves agentic AI systems composed of multiple autonomous, collaborative computational agents capable of planning, reasoning, retrieving information, conducting analyses, evaluating uncertainty, adapting to changing environments, and coordinating complex tasks with limited human intervention. These systems resemble distributed intelligence teams in both architecture and function.

Within a NEURINT enterprise, multiple specialized agents could concomitantly evaluate distinct dimensions of human cognition. As illustration, one agent could analyze political discourse, while another models leadership psychology. Another agent evaluates cultural narratives, while another examines social media sentiment. Progressing farther into bio-psychosocial dimensions, one agent would integrate neuroscientific evidence regarding stress, threat perception, or decision biases, while additional agents continually compare these outputs against historical precedents and evolving intelligence. A supervisory orchestration layer could synthesize these functions and analyses into coherent assessments that dynamically evolve as additional information becomes available. Such architectures could increase analytical depth while reducing latency between data acquisition and actionable intelligence.

Big Data as the Substrate of Neurocognitive Intelligence

NEURINT is fundamentally data intensive, its effectiveness dependent upon integrating extraordinarily diverse information streams, as defined in Table 2. 

Table 2
Types of Data and Information Engaged in/by NEURINT

• Traditional intelligence reporting
• Open-source information
• Multilingual media
• Scientific literature
• Technological innovation metrics
• Demographic datasets
• Economic indicators
• Behavioral analytics
• Social network structures
• Communications metadata
• Commercial sensing
• Epidemiological trends
• Historical databases

Importantly, these data are semantically integrated across disparate domains, such that previously cryptic relationships would become computationally detectable and patterns of influence emerge. To reiterate, and perhaps most importantly, these relationships evolve continuously as new data enter the system, and in this way, NEURINT can be regarded as a progressively developing and maturing analytical intelligence ecosystem. 

Constructing Cognitive Digital Twins

Among the most transformative applications of agentic AI is the creation of computational cognitive digital twins, which are defined as probabilistic computational representations of individuals, organizations, and/or populations constructed from multiple behavioral and informational inputs. Cognitive twins represent adaptive models that continually update as new information becomes available such that variables within computational models become capable of forecasting likely responses under varying conditions. Such modeling produces probability distributions that can assist analyses important to identifying decision pathways, escalation thresholds, vulnerabilities, and opportunities for influence. Notably, the objective is not perfect prediction, but fortified strategic anticipation.

But contemporary conflict increasingly involves networks of individuals operating in varying degrees of collaboration, cooperation, competition and conflict. Insurgent organizations, extremist movements, biotechnology collaborations, cyber collectives, and state-sponsored influence operations all emerge from distributed human systems. When operationalized within NEURINT architecture(s), agentic AI can enable simultaneous modeling of multiple interacting cognitive networks to characterize a number of factors and trends, as shown in Table 3. 

Table 3
Factors and Trends Assessed and Analyzed by Agentic AI within NEURINT Architecture(s)

• Leadership cohesion
• Internal dissent
• Ideological fragmentation
• Organizational resilience
• Recruitment dynamics
• Innovation networks
• Communication bottlenecks
• Trust relationships
•Narrative propagation

These analyses permit identification of leverage points where relatively modest interventions may produce disproportionately large strategic effects. We posit that such understanding can directly support left-of-bang deterrence.

From Prediction Toward Anticipatory Intelligence: NEURINT and Left-of-Bang Deterrence

The NEURINT model is aimed at facilitating anticipation. The conjoinment of generative agentic AI can enable probabilistically developed alternative future scenarios based upon iterative behavioral models in which 1) multiple competing hypotheses can simultaneously evaluated; 2) counterfactual simulations can model how differing events could influence adversary cognition; 3) each can be computationally evaluated before and during occurrence, and 4) the resulting forecasts are therefore iteratively weighed (against prior probabilistic inference(s)). Taken together, this paradigm can afford insight(s) to differentially evaluated and validated strategic options in advance of crises emerging.

Left-of-bang deterrence (i.e., prior to engagement) seeks to prevent hostile capability from becoming operational, characteristically by logistics, communications, constraining financial support, and operational planning that disrupt adversaries’ weapons development. NEURINT can expand this concept to addresses cognitive systems such that preemptive deterrence becomes directed toward preventing hostile intent, fracturing organizational cohesion; redirecting innovation pathways, raising anticipated weapon development and deployment costs and altering adversarial decision calculi. These objectives may often be achieved non-kinetically.

This framework may be particularly relevant to weaponizable bioscience and biotechnology. The rapid democratization of synthetic biology, AI-enabled molecular engineering, autonomous laboratory systems, and global scientific collaboration have all altered biological threat landscapes. To be sure, traditional intelligence may identify laboratories, materials, personnel, and/or financial support. However, NEURINT could add another dimension by affording insights to why actors pursue particular biological capabilities, what incentives motivate innovation; what organizational cultures accelerate risk-taking, and which variables and leadership dynamics encourage escalation?

Toward addressing these questions, agentic AI can integrate scientific publication networks, patent databases, funding structures, conference participation, commercial biotechnology development, social media interactions, educational exchanges, and geopolitical events into continuously evolving assessments of dual-use innovation ecosystems. Such insight could provide valuable support for interventions that disrupt dangerous developmental trajectories before weaponization occurs.

Human-Machine Cognitive Teaming: Functional Reciprocity and Ethical Responsibility

NEURINT should never become fully automated, not simply in rote adherence to Department of Defense Directive 3000.09, but as reflective of the Directive’s grounding recognition that human judgment remains indispensable, and human accountability persists as the bedrock of operational responsibility. Experienced ISR operators contribute contextual understanding, ethical reasoning, cultural insight, strategic intuition, and political awareness that exceed the capabilities of current agentic AI. Accordingly, NEURINT should be conceived as a human-machine system partnership in which agentic AI engages continuous monitoring, large-scale data integration, pattern recognition, scenario generation, and uncertainty estimation. Complementarily, human analysts provide strategic interpretation, information and implication validation, decisional support, operational prioritization, ethical oversight, and policy alignment and concurrence. The objective is to develop and facilitate truly synergistic cognitive systems for ISR.

Merely sustaining what we refer to as human “loop presence “ (e.g., defining how  human agents can and should be “near the loop,” “on the loop,” or “in the loop” of the human-machine system dyad) does not guarantee that the system will function and/or be employed with ethical probity. Indeed, any such exercise of modeling, assessment and analytics can bear inherent risks of bias, overconfidence, inappropriate surveillance, and misuse. Accordingly, we opine that the following principles should guide operational development and deployment. First, transparency regarding analytical assumptions should be maximized whenever operational security permits. Second, uncertainty estimates should accompany all computational assessments. Third, multiple competing models should routinely challenge prevailing conclusions. Fourth, human review must remain integral to consequential decisions; and lastly, ethical norms, legal authorities, constitutional protections, and international obligations must inform all operational use-in-practice. The objective is to enhance democratic security, not compromise or impugn it. 

Recommendations

To operationalize AI-enabled NEURINT, the following recommendations are offered: 

1. Creation of an interagency NEURINT Center of Excellence to integrate neuroscience, behavioral science, computational social science, AI engineering, intelligence analysis, ethics, and operational planning.

2. Development of secure agentic AI architectures capable of continuously integrating classified and appropriately authorized open-source datasets while preserving provenance, explainability, and analytic confidence metrics.

3. Construction of validated cognitive digital twin frameworks of leadership teams, strategic organizations, technological ecosystems, and relevant adversarial networks that can be used to support anticipatory assessment and strategic gaming.

4. Integration of NEURINT capabilities and products directly into left-of-bang planning for biodefense, cyber operations, influence campaigns, counterterrorism, counterproliferation, and strategic competition, ensuring that cognitive indicators complement—not replace—traditional intelligence disciplines.

5. Establishment of governance frameworks that incorporate legal review, neuroethical guidance, independent validation, and continuous red-teaming to ensure that AI-enabled neurocognitive intelligence remains effective, trustworthy, and aligned with democratic values.

Conclusion

Strategic competition is increasingly a contest over cognition, and success will be dependent upon understanding how key technologies interact with human perception, judgment, emotion, and decision-making. Our earlier conception of NEURINT anticipated this technical evolution by proposing that ISR operations and methods must increasingly focus upon characterizing the neural and cognitive determinants of behavior. Our initial construct of NEURINT can now be operationalized through the convergence of generative agentic AI, big data analytics, computational neuroscience, and advanced behavioral modeling.

Properly guided and governed, we submit that NEURINT could constitute a new integrative layer capable of fusing disparate sources of information into dynamic assessments of the cognitive battlespace. Through adaptive cognitive modeling, digital twin architectures, probabilistic forecasting, and human-AI cognitive teaming, NEURINT could provide earlier warning, deeper insight, and more nuanced comprehension of the motivations, vulnerabilities, and decision processes that underlie strategic behavior.

Most importantly, NEURINT could shift ISR from reactive observation toward anticipatory engagement. By elucidating the cognitive conditions from which hostile intent emerges, it can enable operational and policy leaders to engage variables to shape environments before competitive crises mature into frank conflict. In this regard, AI-enabled NEURINT may be particularly well suited to left-of-bang deterrence, wherein success is measured not by post-attack retaliatory measures (i.e., deterrence by denial), but by preventing or at least mitigating threat trajectories from achieving operational fruition. Thus, as biotechnologic innovation, AI, and information ecosystems become increasingly interdigitated, the ability to understand and aptly influence the cognitive foundations of strategic behavior will become an indispensable national security asset. AI-enabled NEURINT may offer a viable path and potentially valuable toolkit toward achieving that objective.

Disclaimer

The views and opinions expressed in this essay are those of the authors and do not necessarily reflect those of the United States government, Department of War, or the National Defense University.

Dr. Elise Annett is a Research Fellow in the Program for Disruptive Technology and Future Warfare of the Institute for National Strategic Studies at the National Defense University. 

Dr. James Giordano is Head of the Center for Strategic Deterrence and Study of Weapons of Mass Destruction, and Program Lead for Disruptive Technology and Future Warfare of the Institute for National Strategic Studies at the National Defense University.