AI should learn the unsaid: rather than only processing explicit knowledge, future systems must model experts' tacit sensing longitudinally through a Tacit Signal Infrastructure and Long-term Cognitive Operations. To make this practical and accountable the paper proposes a new professional role—the Cognitive Operations Manager—and a training/research framework (CORTF) to govern, validate, and train these capabilities.

Main Finding

The paper argues that next‑generation AI should move beyond processing explicit knowledge and instead model experts’ tacit sensing over time. It introduces Tacit Signal Infrastructure (TSI) — a cognitive infrastructure layer to capture, structure, model, and validate weak/tacit signals longitudinally — and defines Long‑term Cognitive Operations (LTCO) as the operational practices needed to maintain and govern such systems. It proposes a new AI‑native professional role (Cognitive Operations Manager) and a research/training architecture (CORTF) to operationalise this vision.

Key Points

• Limit of current systems: Generative AI excels at explicit knowledge tasks (retrieval, summarisation, codified workflows) but lacks capacity to model high‑level expert cognition that relies on tacit sensing (weak signals, coherence/tension detection, anticipatory judgement).
• Tacit sensing: Experts perceive weak, ambiguous, temporally unfolding cues (e.g., rhythm disruption, semantic instability, emergent tension) that typically are not captured in formal logs or short interactions.
• Tacit Signal Infrastructure (TSI): A proposed infrastructural layer to:
  • Capture longitudinal cognitive traces and weak signals across artefacts, interactions, and contexts.
  • Structure and semantically organise those traces (memory curation, semantic networks).
  • Model weak signals without forcing reduction to fixed rules (longitudinal calibration, probabilistic/attentional models).
  • Support expert review, contestability, and governance.
• Long‑term Cognitive Operations (LTCO): Operational practices required to run TSI:
  • Memory curation (what historical traces to keep, prune, timestamp).
  • Semantic organisation (ontologies, evolving taxonomies, coherence metrics).
  • Tacit signal modelling (feature engineering for weak signals, temporal models, interpretability).
  • Reasoning calibration (human-in-the-loop validation, calibration of model sensitivity/precision).
  • Cognitive governance (ethics, contestability, role definitions, auditability).
• New professional capabilities: The Cognitive Operations Manager (COM) as a prototype role to coordinate signal capture, semantic modelling, calibration, expert validation, and governance.
• Research & training: Cognitive Operations Research & Training Framework (CORTF) to support methods, curricula, and workforce development for LTCO and TSI design.
• Conceptual foundations: The vision synthesises literature on tacit knowledge, weak signals, distributed/extended cognition, infrastructure studies, HCI, and organisational learning.

Data & Methods

• Nature: Vision/conceptual paper (no empirical dataset or experimental results).
• Methodology: Interdisciplinary theoretical synthesis and conceptual framing drawing on:
  • Tacit knowledge and expertise literature (Polanyi; Dreyfus; Klein; Schön).
  • Weak signals, foresight, early warning research.
  • Human–AI interaction, sensemaking, tacit elicitation, and HCI studies.
  • Distributed cognition and infrastructure scholarship.
• Outputs: Definitions of new constructs (Tacit Signal Infrastructure, Long‑term Cognitive Operations), an operational role (COM), and a training/research architecture (CORTF). The paper proposes design principles and research directions rather than empirical validation.

Implications for AI Economics

• Labour demand and skill composition:
  • Creates demand for new AI‑native, high‑cognitive roles (Cognitive Operations Managers, LTCO specialists) with a mix of domain expertise, ML/AI systems knowledge, and governance skills.
  • Changes complementarity patterns: AI systems that model tacit signals are complements to experienced experts rather than simple substitutes, potentially raising returns to firms that combine AI with deep domain expertise.
  • Potential for both upskilling (new professional trajectories) and deskilling (if tacit tasks are automated poorly); net effect depends on technology design and governance.
• Productivity and value capture:
  • TSI could raise productivity in domains where early detection yields high value (finance, healthcare, project management, high‑reliability operations) by shortening time‑to‑detection or enabling earlier corrective action.
  • Firms that invest early in TSI and LTCO capabilities may capture disproportionate rents due to knowledge capital, path dependency, and lock‑in (historical traces + curated semantic models).
• Investment and adoption incentives:
  • Building TSI is infrastructure‑heavy (data curation, long‑horizon storage, governance), creating large fixed costs and increasing economies of scale / scope. This favors incumbents and large firms with resources to sustain long‑term operations.
  • Unclear short‑term ROI and measurement challenges (benefits are anticipatory and probabilistic) may slow adoption among smaller organisations; service providers may offer TSI-as-a-service.
• Market structure and competition:
  • Potential concentration: richer actors who can afford to curate long histories and hire COMs can develop superior tacit models, raising barriers to entry.
  • New markets: tools and consultancies for tacit signal capture, LTCO platform services, standards/audit providers, and specialised training providers.
• Measurement and empirical challenges:
  • Standard productivity metrics and short‑horizon evaluations understate TSI benefits because gains are realized through earlier warnings and avoided tail events rather than immediate throughput increases.
  • Need for new performance metrics: early‑warning lead time, true/false positive costs, calibration of human‑AI trust, downstream avoided losses, and survival/continuity metrics.
• Risk, governance, and externalities:
  • Miscalibration risks: false alarms (costly interventions) and missed signals (systemic failures) create asymmetric costs needing explicit calibration and human adjudication.
  • Liability and moral hazard: reliance on TSI outputs changes accountability chains; firms may shift responsibility to models without adequate governance.
  • Surveillance and privacy concerns: long‑horizon capture of tacit traces (communications, behavioural cues) raises legal/ethical issues and potential regulatory intervention costs.
• Policy and regulatory implications:
  • Standards and audit frameworks will be needed for TSI (data retention policies, interpretability requirements, validation protocols).
  • Workforce policy: need for training subsidies, accreditation for COMs, and updated labor market forecasts to manage transitions.
• Empirical research agenda for economists:
  • Firm‑level studies: difference‑in‑differences around TSI/LTCO adoption to estimate productivity, retention, error reduction, and risk mitigation effects.
  • Field experiments: randomized rollouts of tacit‑signal alerts in domains like healthcare triage or project management to measure decision changes and outcomes.
  • Labor market analyses: wage premia for COM‑type roles, reallocation across occupations, and education/training returns.
  • Market structure analyses: examine whether infrastructure investments lead to increased concentration and whether TSI‑providers act as multipliers or gatekeepers.
  • Cost‑benefit modeling: quantify trade‑offs of storage/curation costs vs. avoided tail risks and improved anticipatory performance.
• Strategic guidance for firms and investors:
  • Invest in pilot LTCO processes and measurable use‑cases with clear counterfactuals (e.g., earlier detection of project failure).
  • Treat tacit signal capture as long‑term strategic infrastructure — budget for storage, curation, human validation, and governance.
  • Combine investments in domain expertise with AI engineering and governance to maximise complementarity and reduce miscalibration risk.
• Potential macro effects:
  • If widely adopted in high‑impact sectors, TSI could reduce frequency/severity of some systemic shocks (faster detection), but could also increase correlated behaviour (herding) if many actors follow similar tacit models — creating new systemic risk channels.

Concise research/practical next steps - Empirical pilots in domains with clear downstream costs of late detection (healthcare adverse events, major projects, supply chains). - Define measurable metrics (lead time, cost of false positives/negatives, avoided loss) and run controlled evaluations. - Develop curricula and certification pilots for Cognitive Operations Managers. - Study market dynamics of TSI providers and potential regulatory frameworks to balance innovation, accountability, and privacy.