The proliferation of generative AI has created a crisis that no existing measurement instrument captures. Humans are increasingly present in knowledge systems without being cognitively engaged. They occupy nodes in information networks, receiving AI-generated outputs and forwarding them onward, but the cognitive work that once occurred at those nodes (questioning, synthesising, judging, generating) has been offloaded to machines. The humans remain in the loop, but the loop has become a conveyor belt.

The Cognitive Vitality Index™ (CVI) measures the proportion and quality of genuinely human-engaged cognitive work within a knowledge system, at any scale.

It is grounded in a single principle: cognitive capabilities that are not practised erode over time. Below a critical threshold of genuine engagement, humans transition from active thinkers to passive routers of information, becoming functionally indistinguishable from automated agents. The CVI tracks this threshold across multiple dimensions, providing both a snapshot score and a trajectory over time.

The Use-It-or-Lose-It Thesis

The neuroscience is unambiguous. [1] Cognitive reserve, the brain’s accumulated resilience built through sustained intellectual engagement, operates on a use-it-or-lose-it principle. Capabilities maintained through regular practice resist degradation; capabilities allowed to lie fallow atrophy.

Aviation provides the longest-studied analogue. [2] Pilots in highly automated cockpits lose manual flying skills because automation handles the cognitive work that once kept those skills sharp. When automation fails and manual capability is suddenly required, the degradation becomes lethal. The parallel to knowledge work is direct: professionals who offload critical thinking to AI lose the capacity for critical thinking when AI is unavailable, unreliable, or wrong.

The FAA has documented this pattern formally [3] : decision-support automation, intended to enhance human performance, instead degrades the very capabilities it was designed to augment. The effect is not immediate. It accumulates over months and years of practice avoidance, making it invisible until a crisis reveals the erosion.

Medicine confirms the same trajectory. [4] Diagnosticians who rely on AI decision support show measurable declines in diagnostic accuracy when working independently. The AI did not make them better doctors; it made them dependent on assistance that may not always be available, appropriate, or correct.

The crisis is not that AI is inadequate. The crisis is that humans stop practising the cognitive work that keeps them capable of evaluating whether AI is adequate.

What “Genuinely Engaged” Means

The CVI distinguishes between two modes of human participation in cognitive work:

Transformation (vital engagement): the human is questioning inputs, synthesising across sources, applying judgment and values, generating novel framings, identifying what is missing, and overriding recommendations when appropriate. The information that exits the human node is qualitatively different from what entered it.

Pass-through (eroded engagement): the human accepts inputs at face value, forwards single sources without synthesis, applies rules mechanically without judgment, repackages in standard formats without critique, processes only what is presented without questioning gaps, and defaults to AI recommendations without evaluation. The information exits the human node essentially unchanged.

Vital (Transformation)Eroded (Pass-through)
Questioning the inputAccepting at face value
Synthesising across sourcesForwarding a single source
Applying judgment and valuesApplying rules mechanically
Generating novel framingsRepackaging in standard formats
Identifying what is missingProcessing only what is presented
Overriding when appropriateDefaulting to AI recommendation
A human in permanent pass-through mode is no longer a thinker. They are infrastructure.1 A human in permanent pass-through mode is no longer a thinker. They are infrastructure.1

Composability Across Scale

The CVI is designed as a composable metric2 composable metric2 . The same definition applies at every level of analysis; what changes is the additional network-flow dimensions that emerge at the collective level.

Individual level (atomic unit): What proportion of this person’s cognitive work involves genuine transformation versus pass-through? Measured through interaction patterns, override frequency, synthesis quality, and independent problem-solving capability when AI is unavailable.

Collective level (network property): What proportion of information flowing through this system passes through genuinely engaged human cognition? At this scale, additional dimensions emerge: how information routes between nodes, whether certain humans become bottlenecks of genuine engagement while others become pure pass-through relays, and whether the system can maintain function when any single high-engagement node is removed.

The collective CVI is not simply an average of individual scores. [5] It is a network property that accounts for flow volume, node criticality, and redundancy of genuine engagement across the system. An organisation could have high average individual CVI but dangerously concentrated engagement (all genuine thinking occurs in three people out of a hundred), or moderate individual CVI distributed robustly across all nodes.

Dimensions of Cognitive Vitality

The CVI decomposes into major categories visualised on a radar chart for immediate high-glance value. These categories are not independent variables; they interact in complex ways documented through the companion Resonance Wheel3 Resonance Wheel3 .

The full taxonomy operates as a hierarchy: six major categories, each decomposing into sub-dimensions and measurable attributes. The major categories (subject to empirical refinement through Design Science Research) include:

  • Agency Retention: Frequency and quality of human-initiated overrides, modifications, and independent decisions
  • Critical Thinking Maintenance: Capacity for analytical reasoning, evidence evaluation, and hypothesis generation without AI support
  • Skill Preservation: Competency in domain-specific tasks performed independently of AI assistance
  • Epistemic Autonomy: Perceived and actual ownership over knowledge generation and truth verification
  • Cognitive Load Balance: Ratio of generative thinking to verification and monitoring of AI outputs
  • Synthesis Quality: Ability to integrate information across sources into novel framings that did not exist in any single input

At the individual level, these dimensions are measurable. [6] Wu et al. (2025) have already validated a 22-item psychometric scale for AI dependence with four distinct dimensions, demonstrating that these phenomena can be captured through rigorous instrumentation.

At the collective level, additional dimensions emerge:

  • Engagement Distribution: How concentrated or distributed is genuine cognitive engagement across the network?
  • Flow Transformation Ratio: What percentage of information flows involve human transformation versus direct routing?
  • Resilience Under Absence: Can the system maintain cognitive output quality when high-engagement nodes are temporarily removed?

The Critical Threshold

The CVI is not merely descriptive. It operationalises a warning system. [7] The thesis, supported by evidence from aviation, medicine, and neuroscience, is that cognitive capabilities exhibit non-linear degradation. Performance maintains at an acceptable level during gradual disuse, then drops sharply below a critical threshold of practice frequency.

The mechanism of erosion is habituation. [8] Mladin et al. (2026) document how users progressively adapt to the predictive regularities of AI systems through “algorithmic habituation,” a process that enhances efficiency in the short term but erodes initiative, creativity, and independent judgment over time. The four dimensions of habituation (cognitive, decisional, creative, moral) map directly onto the CVI dimensions they degrade.

The CVI tracks trajectory toward or away from this threshold. A declining CVI is an early warning that cognitive capabilities are entering a degradation spiral. The intervention point is before the threshold is crossed, not after capabilities have already eroded beyond recovery.

Relationship to the Perceptiosphere

The Perceptiosphere defines the architecture of knowledge sovereignty: nested zones of cognitive influence within which knowledge is created, curated, and transmitted. The CVI measures the health of human cognition operating within that architecture.

These are complementary but independent constructs:

  • A well-designed Perceptiosphere with low CVI: the architecture exists, but humans have stopped genuinely engaging with it. The structure is sound; the inhabitants are hollow.
  • High CVI within a poorly designed knowledge system: engaged humans fighting bad architecture. Unsustainable but temporarily functional.
The ideal state is high CVI within a well-designed Perceptiosphere: humans are genuinely engaged in cognitive work, and the architecture supports and rewards that engagement rather than incentivising offloading.

Visual Companions

The CVI is communicated through two paired visualisations:

Radar chart: Plots the major dimensions on polar coordinates, providing immediate high-glance value. Excellent for comparing two snapshots over time (trajectory visualisation) and for communicating the overall “shape” of cognitive health.

Resonance Wheel™: [9] A chord-diagram-based companion that reveals the interdependencies between dimensions. Shows which dimensions “resonate” when one is changed, making trade-offs and cascading effects visible. Corrects the reductionist illusion inherent in any radar chart.

Together, they answer two questions: “What is the current state?” (radar chart) and “What happens if we try to change it?” (Resonance Wheel).