Learning Objectives
- Translate AI and algorithmic systems instability into P, ΔV, σ, and Lr.
- Identify how AI Systems signals become legally relevant before visible failure.
- Apply asymmetric uncertainty treatment to AI Systems data.
- Calculate prudential implications for Safe Mode, Restoration First, and CSAM escalation.
- Convert sector data into c-ECO contractual and institutional consequences.
The Threshold Function Protocol in AI Systems
AI & Algorithmic Systems systems are threshold-sensitive because ordinary continuity can conceal progressive loss of reversibility. Module 2 translates sector facts into the four TFP variables and teaches Fellows to distinguish measurement, interpretation, and governance consequence.
The AI Systems trigger classification is a function of position, trajectory, uncertainty, and reversibility liquidity.
Sector Calibration Principle
The variables remain stable across c-ECO. What changes is empirical content. In this track, calibration begins with models, training data, automated decisions, computational infrastructure, human oversight, governance controls, public impact, and institutional accountability. Fellows must define which system is protected, which threshold matters, which signals are decision-grade, and which interventions remain reversible.
The Four TFP Variables in AI Systems
Definition: The current state of an activity, asset, environment, or system within its systemic stability space, measured relative to relevant thresholds, Safe Operating Space boundaries, and potential failure conditions.
AI Systems translation: P is assessed through acceptable error, drift, and harm boundaries, human oversight capacity limits, data integrity and bias thresholds, and through the proximity of the case to operational, ecological, social, or institutional failure.
Low P does not mean harm has occurred. It means the system is close enough to a relevant boundary that ordinary continuation assumptions must be challenged.
Definition: ΔV measures whether the system is moving toward or away from threshold conditions, and how quickly.
AI Systems translation: Fellows examine model drift, performance degradation, and unexplained decision variance, data bias, representational failure, and dataset obsolescence, compute and energy burden acceleration. Sustained negative velocity may justify intervention even before a formal boundary is crossed.
Definition: σ captures sensor error, incomplete monitoring, model limitations, data discontinuity, institutional blind spots, and contested evidence.
Critical principle: In c-ECO, uncertainty does not create permission to ignore deteriorating trajectories. Where reversibility is shrinking, uncertainty narrows the acceptable margin.
Definition: Lr measures whether immediately mobilizable resources, institutional authority, technical options, and time remain sufficient to stabilize or redirect the case.
AI Systems translation: Rmi may include enforceable funding, technical capacity, substitution options, emergency authority, monitoring access, and contractual leverage. Ct is the projected cost of stabilization, redesign, or recovery.
Sector Signal Library
| Signal | TFP Use | Governance Question |
|---|---|---|
| Model drift, performance degradation, and unexplained decision variance | P proximity | Does this signal show that the AI Systems case is stabilizing, degrading, or approaching a critical decision boundary? |
| Data bias, representational failure, and dataset obsolescence | ΔV direction | Does this signal show that the AI Systems case is stabilizing, degrading, or approaching a critical decision boundary? |
| Compute and energy burden acceleration | σ weighting | Does this signal show that the AI Systems case is stabilizing, degrading, or approaching a critical decision boundary? |
| Automation dependency and human oversight erosion | Lr pressure | Does this signal show that the AI Systems case is stabilizing, degrading, or approaching a critical decision boundary? |
| Adverse decision clustering, externality concentration, and public trust loss | Safe Mode relevance | Does this signal show that the AI Systems case is stabilizing, degrading, or approaching a critical decision boundary? |
Problem Set: Variable Calibration
Scenario: An automated decision system, model deployment, AI infrastructure dependency, algorithmic governance process, or compute-intensive platform with systemic risk or public-interest exposure.
Tasks: Define the system boundary; identify direct and indirect actors; state which SOS boundary or failure condition is most relevant; explain what would make the case unsuitable for CSAM development.
Choose two signals from the sector signal library. Assign a plausible current state, reference range, and boundary. Calculate a nominal P and describe whether ΔV is improving, stable, or deteriorating.
Identify three evidence gaps. Explain whether they increase σ, reduce Lr, or both. Draft one immediate information request and one reversible intervention option.
Compare three assets, territories, contracts, or institutional units inside the same AI Systems system. Rank them by systemic urgency and justify the ranking through P, ΔV, σ, and Lr.
Draft a two-page CSAM technical annex identifying variables, evidence sources, monitoring frequency, threshold assumptions, and the first point at which institutional escalation becomes justified.
Preparation Guide
Step 1 — 90 min: Revisit Module 1 Key Concepts and the TFP preview. Identify how P and ΔV differ in your selected case.
Step 2 — 90 min: Gather public or cohort-provided data on model drift, performance degradation, and unexplained decision variance, data bias, representational failure, and dataset obsolescence, compute and energy burden acceleration.
Step 3 — 120 min: Complete Problem Set A with explicit assumptions and uncertainty notes.
Step 4 — 90 min: Draft a one-page memo: When does AI and algorithmic systems continuation become incompatible with reversibility?
Required Materials
Primary c-ECO Materials
- TFP Manual sections on P, ΔV, σ, Lr, prudential classification, and Safe Mode conduct.
- Module 1 doctrine: Safe Operating Space, Physical Primacy, Contracting Reversibility, and CSAM formation.
- Fellowship instruments governing methodological fidelity, confidentiality, and cohort submission.
Sector References
- NIST AI Risk Management Framework.
- OECD AI principles.
- ISO AI governance materials.
- EU AI Act risk taxonomy materials.
Assessment
| Component | Weight | Standard |
|---|---|---|
| Problem Set A | 35% | Correct variable definitions, transparent assumptions, and sector-specific measurement logic. |
| Problem Set B | 25% | Comparative ranking demonstrates systemic reasoning rather than ordinary risk scoring. |
| CSAM Annex | 25% | Evidence sources, threshold assumptions, uncertainty, and intervention implications are coherent. |
| Workshop Participation | 15% | Contributes disciplined questions and identifies where data gaps alter governance consequences. |