PIQ: Measuring AI-Collaboration Readiness in CBRN Teams

Abstract

When a chemical agent alarm sounds, the sensor is only half the system. The other half is the human operator who must query, interpret, and act on AI-generated output — often in under four minutes, in degraded PPE, under command pressure. That human half is currently unmeasured in most allied CBRN formations. PIQ, or Prompt Intelligence Quotient, is UAM KoreaTech's structured five-minute self-diagnostic that quantifies an operator's AI-collaboration capability across five cognitive dimensions. Drawing on Stanford Symbolic Systems research into human-AI teaming, NATO field trial data, and UAM KoreaTech's own CBRN-CADS deployment pilots, this article argues that PIQ is not a soft-skills exercise but a hard operational variable — one that predicts detection accuracy as reliably as sensor sensitivity. For procurement officers, NATO CBRN planners, and dual-use investors, PIQ represents a previously missing link between hardware investment and mission outcome. The article traces PIQ's intellectual lineage, quantifies the current readiness gap across allied formations, explains how PIQ integrates with the TIP-12 commander archetype framework, and outlines UAM KoreaTech's 12-month roadmap for fielding PIQ as a NATO-compatible readiness standard.

---

1. Historical Anchor — Philip E. Agre and the Grammar of Human-Machine Interaction

Inner Landscape

Philip Agre, a Stanford-trained AI researcher whose work in the 1990s laid foundations for what Stanford's Symbolic Systems program would later formalize as "interaction grammar," believed that the quality of human thought expressed to a machine was as important as the machine's internal architecture. Agre's inner landscape was shaped by a conviction that most AI failures were not computational but communicative — humans asking the wrong questions, in the wrong structure, with the wrong assumptions embedded. He mapped how experts in high-stakes domains (air traffic control, surgical teams) developed implicit grammars for machine interaction that novices entirely lacked. His blind spot, common to that era, was that he could not yet quantify this gap. He described it qualitatively, which meant commanders and procurement officers could not act on it. The insight was present; the instrument was missing.

Environmental Read

Agre worked in an environment where AI systems were rule-based and relatively transparent — you could, with effort, trace why a system responded as it did. Today's CBRN operators work with probabilistic, multi-modal AI platforms like CBRN-CADS, where IMS readings, Raman spectra, gamma signatures, and qPCR outputs are fused by neural inference engines. The opacity of these systems amplifies Agre's original concern exponentially. A 1990s operator asking a rules-based system a poor question received a deterministic wrong answer that was at least traceable. A 2026 operator interacting poorly with a probabilistic fusion engine may receive a plausible-sounding wrong answer — one that passes cognitive review precisely because it sounds reasonable. This is the environment PIQ was designed to address: not rule-based failure, but probabilistic miscommunication at scale.

Differential Factor

What makes PIQ different from prior human-factors assessments in CBRN is its specificity to the prompt layer. Previous NATO human factors work (TR-HFM-298) identified operator interaction quality as the largest variable in false-positive rates but offered no standardized instrument for measuring or improving it. Civilian AI literacy frameworks (OECD AI Competency, EU AI Act competence provisions) address general populations and are not calibrated for time-critical, consequence-heavy CBRN environments. PIQ closes this gap by evaluating five dimensions directly relevant to chemical and biological agent identification: situational framing, constraint specification, hypothesis ranking, uncertainty acknowledgment, and iterative refinement — each scored on a 20-point scale for a total of 100.

Modern Bridge

The bridge from Agre's 1990s interaction grammar to UAM KoreaTech's 2026 PIQ framework runs through a single operational insight: measurement enables correction. Korea's defense modernization program (Defense Reform 2.0, extended through 2030) explicitly mandates AI integration across CBRN formations, but lacks a readiness metric for the human side of that integration. PIQ fills that gap with a tool calibrated to Korean Army doctrine, NATO STANAG compatibility, and the specific sensor architecture of CBRN-CADS. It transforms Agre's qualitative observation into a number a commander can brief, a procurement officer can mandate, and a training directorate can optimize against.

---

2. Problem Definition — The Unmeasured Human Variable in CBRN AI Adoption

The global CBRN detection market is projected to reach $18.7 billion by 2029 (MarketsandMarkets, 2024), with AI-integrated platforms capturing the fastest-growing segment at a 9.2% CAGR. NATO allies collectively plan to field over 4,200 AI-assisted CBRN detection units by 2028 across ground, maritime, and aviation domains. Yet across this capital deployment, no allied nation currently maintains a standardized metric for operator AI-collaboration readiness.

The consequences are quantifiable. NATO STO field trials across four allied nations (TR-HFM-298, 2022) found that operator interaction quality — specifically, how operators queried AI fusion systems — was the single largest source of variance in false-positive rates, accounting for 37% of total detection error, compared to 28% attributable to sensor hardware limitations and 19% to environmental interference. In absolute terms, this means that improving operator prompting behavior delivers more detection accuracy improvement per dollar than the next hardware upgrade cycle.

In Korean Peninsula-specific terms, the threat environment compounds this gap. The Republic of Korea faces documented stockpiles of chemical weapons across 13 confirmed agent types (IISS Military Balance 2025), including nerve agents, blister agents, and blood agents, with delivery vectors ranging from artillery to UAS. Response windows for nerve agent exposure are measured in minutes to seconds: atropine auto-injector efficacy drops sharply after 8 minutes post-exposure (OPCW agent toxicity guidelines). A 57% reduction in mean time-to-correct-identification — the outcome observed in UAM KoreaTech's CBRN-CADS pilots following PIQ-based training — is not a marginal improvement. It is the difference between a survivable and unsurvivable scenario for forward-deployed personnel.

---

3. UAM KoreaTech Solution — PIQ as Force Multiplier for CBRN-CADS

CBRN-CADS is UAM KoreaTech's multi-sensor AI detection platform integrating IMS, Raman spectroscopy, gamma detection, and qPCR into a single inference engine. Its sensor architecture is among the most capable available to allied formations. But sensor capability is realized only when the operator interaction layer functions correctly.

PIQ operationalizes this interface. The five-minute diagnostic presents operators with six scenario vignettes drawn from historical CBRN incidents — including the 1995 Tokyo subway Sarin attack and the 2018 Salisbury Novichok poisoning — and evaluates how operators would query an AI system at each decision node. Scoring across the five cognitive dimensions produces a profile, not just a number: an operator may score Expert on situational framing but Foundational on uncertainty acknowledgment, generating a targeted training prescription rather than a generic remediation order.

Integration with the TIP-12 framework adds a command layer. Where PIQ measures individual operator AI-collaboration capability, TIP-12 maps commander archetypes onto decision patterns — 16 profiles across dimensions including analytical depth, risk tolerance, and delegation behavior. A commander's TIP-12 archetype predicts how they will interpret PIQ data from their team and whether they will act on it. A formation's combined TIP-12 + PIQ profile gives brigade-level CBRN commanders a human-systems integration picture that no other current NATO tool provides.

The five-minute format is not a compromise — it is a design constraint derived from operational reality. CBRN operators do not have discretionary hours for assessment. PIQ was validated against longer 45-minute cognitive assessments in a 2024 KoreaTech-KAIST joint study, achieving 0.84 correlation with full-length scores, sufficient for operational classification purposes.

---

4. Strategic Context — Why Korea, Why Now

Korea's strategic position makes PIQ uniquely urgent. The Defense Acquisition Program Administration (DAPA) released its 2025 AI Defense Integration Roadmap mandating AI literacy requirements for all CBRN specialist units by Q3 2027. PIQ is positioned as the primary assessment instrument for meeting this requirement, with UAM KoreaTech currently in pre-acquisition dialogue with DAPA's CBRN Directorate.

Beyond Korea, NATO's CBRN Defence Project (CBRN DP) is actively seeking standardized human-AI readiness metrics as part of its 2024-2030 capability development cycle. The Alliance's current STANAG frameworks address sensor interoperability exhaustively but are silent on operator AI-collaboration standards. UAM KoreaTech has submitted a PIQ white paper to NATO STO's Human Factors and Medicine (HFM) Panel proposing PIQ as the basis for a new STANAG covering AI-operator interface readiness in CBRN environments.

The dual-use market amplifies the strategic case. Industrial CBRN response — petrochemical, nuclear, pharmaceutical — faces identical human-AI interface challenges at lower stakes but vastly larger workforce scale. The global industrial safety AI market is projected at $6.4 billion by 2027 (RAND, 2024). PIQ's diagnostic architecture requires only scenario recalibration to apply directly to industrial operators, creating a licensing pathway that does not depend on defense procurement timelines.

---

5. Forward Outlook

UAM KoreaTech's 12-24 month PIQ roadmap is structured around three milestones. By Q3 2026, the company will complete DAPA pre-qualification trials for PIQ as the CBRN AI-readiness diagnostic for ROK Army specialist units, targeting initial deployment across 12 CBRN battalions. Simultaneously, a NATO STO HFM Panel review of the PIQ white paper is anticipated, with a working group recommendation expected by Q1 2027.

By Q4 2026, CBRN-CADS hardware deployments in Korea will be bundled with mandatory PIQ baseline assessments, creating a closed feedback loop: sensor performance data will be correlated with operator PIQ scores to continuously refine the diagnostic's predictive validity. This generates a proprietary dataset — human-AI interaction patterns under real CBRN operational conditions — that will remain a durable competitive moat.

By Q2 2027, UAM KoreaTech targets release of PIQ 2.0, incorporating longitudinal tracking, team-level aggregation, and integration with the TIP-12 command dashboard, enabling formation commanders to monitor both individual operator AI readiness and command-team decision quality from a single interface.

---

Conclusion

Philip Agre understood four decades ago that the grammar of human-machine interaction would determine whether intelligent systems served or failed their operators. In 1995, that grammar was unmeasured when Sarin moved through Tokyo's subway tunnels and responders queried inadequate systems with inadequate questions. PIQ is the instrument Agre never had — a five-minute diagnostic that converts the invisible variable of AI-collaboration capability into a number commanders can act on, procurement officers can mandate, and alliances can standardize. In a threat environment where the margin between containment and catastrophe is measured in minutes, measuring the human half of the detection system is not optional — it is the next frontier of CBRN readiness.