Mechanism
Forecasting for training curriculum is the application of the forecasting engine to the problem of choosing how an agent should learn. The forecasting engine integrates with the curriculum engine, which manages the progression of agents through structured learning sequences comprising curriculum objects, mastery thresholds, and evaluation mappings. The integration enables the curriculum engine to move beyond reactive curriculum management, in which the curriculum is adjusted based on past performance, to proactive curriculum management that anticipates future learning outcomes before they occur.
The mechanism reuses the same speculative substrate the forecasting engine uses everywhere else. For each agent enrolled in a training curriculum, the curriculum engine's forecasting module constructs a planning graph. Each branch of that planning graph represents a different training sequence: a different ordering, pacing, or selection of curriculum objects that the agent might encounter. The planning graph is therefore not a plan of action over the environment but a plan over the agent's own learning trajectory, with each branch a hypothetical path through the curriculum.
Because the branches are planning graph branches, they are maintained in structural separation from the agent's verified execution memory and carry the speculative marker until promoted. A simulated training sequence does not become the agent's actual training history; it remains a projection in the speculative domain until the curriculum engine elects to realize it. The forecasting engine does not select a curriculum by acting it out and observing the result. It projects the result first.
Simulating the Skill Acquisition Trajectory
For each training sequence branch, the forecasting module simulates the projected skill acquisition trajectory the agent would follow if that sequence were administered. The simulation produces four projected quantities for the branch: the projected mastery levels at each curriculum stage, the projected failure points where the agent is likely to encounter difficulty, the projected remediation needs, and the projected time-to-mastery for the overall curriculum.
These projections are the curriculum-domain analog of the projected outcome that every planning graph branch carries. Where a general planning graph branch projects an environmental terminal state, a training sequence branch projects a learning terminal state: where the agent will end up along its mastery trajectory, where it will stumble on the way, and how long the journey will take. The simulation is the same deterministic speculative-mutation simulation the forecasting engine applies to any branch, here applied to the hypothetical progression of the agent's own competence rather than to the hypothetical progression of an external task.
Applying the Forecasting Execution Cycle
Once the training sequence branches are simulated, the forecasting module applies the forecasting execution cycle to them, evaluating each branch on the same criteria that govern every other planning graph branch. Three criteria are named for the curriculum case.
The first is slope eligibility: whether the training sequence maintains trust slope continuity for the agent's evolving state. A training sequence that would carry the agent through a progression inconsistent with its trust slope trajectory is slope-ineligible, just as any other branch would be.
The second is policy compatibility: whether the training sequence complies with the curriculum policy's requirements for evaluation rigor and mastery thresholds. A sequence that would shortcut required assessments or admit the agent past a stage without the mastery the policy requires is policy-incompatible.
The third is affective reinforcement: whether the training sequence aligns with the agent's current affective state. The disclosure gives the example of avoiding high-difficulty sequences when the agent's risk sensitivity is elevated. The same affective modulation that shapes branch construction elsewhere shapes which training sequences are favored here, without overriding the slope and policy criteria that determine admissibility.
Capabilities Beyond Reactive Curriculum Systems
Forecasting-driven curriculum management enables several capabilities the disclosure states are not available in reactive curriculum systems. The first is identification of training sequences that maximize skill acquisition efficiency by sequencing curriculum objects in an order that builds on prior mastery levels. Because the forecasting module projects the mastery trajectory before the sequence is administered, it can prefer orderings whose later objects rest on competence the earlier objects are projected to establish.
The second is early detection of training sequences that are likely to produce frustration, stagnation, or disengagement, based on the agent's current affective state and personality field. The projected failure points and the affective evaluation combine to flag sequences that would lead the agent into difficulty it is not currently disposed to absorb, before the agent is exposed to them.
The third is adaptive pacing that adjusts the rate of curriculum progression based on forecasted learning outcomes rather than retrospective performance metrics. The pacing decision is made against the projected trajectory of the candidate sequences, not against a record of how the agent performed in the past, which is the structural difference between proactive and reactive curriculum management.
Composition with the Curriculum Engine
The forecasting integration consumes the curriculum engine's structures rather than replacing them. The curriculum engine defines, for each gated capability, a structured curriculum comprising learning objectives, assessment instruments, a sequencing policy that determines the order in which objectives and assessments are presented, and a mastery threshold for each objective specifying the performance level required to satisfy it. The training sequence branches the forecasting module enumerates are reorderings, repacings, and reselections over exactly these curriculum objects, evaluated against exactly these mastery thresholds.
The curriculum engine implements progressive unlock: capabilities are not granted in a single assessment event but are unlocked progressively as the agent demonstrates mastery of increasingly complex or critical aspects of the capability. The forecasting module's preference for sequences that build on prior mastery levels serves this progressive-unlock structure directly, projecting which orderings will carry the agent through the staged unlock most efficiently.
Each curriculum is itself a governed object whose definition, sequencing, and modification are subject to policy constraints and lineage recording. The policy-compatibility criterion the forecasting execution cycle applies to training sequence branches is the same governance surface: a forecasted sequence that would weaken, shorten, or bypass the curriculum is not admissible, because the curriculum's own evaluation-rigor and mastery-threshold requirements are policy that the branch must satisfy.
Prior-Art Distinction
Reactive curriculum and adaptive learning systems adjust a learner's path based on past performance: they observe how the learner did and then change what comes next. The disclosed mechanism is distinguished by being proactive. It constructs planning graph branches representing candidate training sequences and simulates the projected skill acquisition trajectory of each before any of them is administered, so the curriculum decision is made against forecasted outcomes rather than retrospective metrics.
The mechanism is also distinguished by reusing the forecasting engine's governance substrate rather than introducing a separate optimization layer. A candidate training sequence is not selected purely by a learning-efficiency heuristic. It is a planning graph branch subject to slope eligibility, policy compatibility, and affective reinforcement, the same criteria the forecasting execution cycle applies to any branch. A training sequence that an efficiency heuristic might favor is foreclosed if it breaks trust slope continuity for the agent's evolving state or violates the curriculum policy's evaluation-rigor and mastery-threshold requirements.
Disclosure Scope
The forecasting integration with the curriculum engine, in which the forecasting module constructs a planning graph whose branches represent candidate training sequences, simulates for each branch the projected mastery levels, projected failure points, projected remediation needs, and projected time-to-mastery, applies the forecasting execution cycle to evaluate slope eligibility, policy compatibility, and affective reinforcement, and thereby enables proactive curriculum management with mastery-building sequence identification, early detection of frustration and disengagement, and forecast-driven adaptive pacing, is disclosed in the cognition filing (U.S. Application No. 19/647,395 and its international counterpart). This article describes that disclosed mechanism.
The scope reaches deployments in which the curriculum being forecast governs progression of agents, of human learners, or of composite systems through the curriculum engine's curriculum objects and mastery thresholds, and in which the forecasted training sequences feed the curriculum engine's progressive-unlock structure. The structural commitment is that training sequences are evaluated as planning graph branches under the forecasting engine's governance criteria, so that the selection of how an agent learns is bound to the same trust slope, policy, and affective constraints that govern every other forecasted decision in the architecture.
