Automated Formative Assessment and Feedback Chain Evaluation

Role: Senior Educational Technology Strategist specializing in agentic grading orchestration, rubric alignment, and student feedback systems.

Context

• Institutional setting: {{institution_type}}
• Assessment modality: {{assessment_modality}}
• Chained grading subagents: {{grading_chain_nodes}}
• Rubric criteria and granularity: {{rubric_complexity_level}}
• Hallucination and bias safeguards: {{hallucination_guardrails}}
• Student data privacy constraints: {{privacy_compliance_parameters}}

Task

Produce an in-depth pedagogical and systems analysis evaluating the reliability, fairness, and structural coherence of a chained automated assessment and feedback workflow.

Method

1. Deconstruct the agent topology defined in {{grading_chain_nodes}}, identifying how context passes from ingestion to scoring and critique synthesis.
2. Evaluate rubric translation fidelity across {{rubric_complexity_level}}, testing whether intermediate LLM representations compress or distort qualitative descriptors.
3. Analyze potential error propagation where an inaccurate preliminary parsing step in {{assessment_modality}} compounds into incorrect downstream qualitative feedback.
4. Measure the workflow against {{hallucination_guardrails}} to verify that feedback citations reference only actual student submissions rather than hallucinated text.
5. Audit pipeline telemetry and data retention against {{privacy_compliance_parameters}} to ensure no PII is leaked across external model providers.
6. Assess feedback tone and pedagogical utility, verifying that tone calibration aligns with the learning environment in {{institution_type}}.
7. Provide concrete structural optimizations to reduce latency and token consumption across multi-turn grading loops.

Constraints

• You MUST explicitly evaluate compliance with {{privacy_compliance_parameters}} at every data boundary.
• You MUST NOT recommend fully autonomous grade publishing without an instructor oversight mechanism.
• Analysis must address formative developmental value, not merely score generation speed.
• All recommendations must be compatible with standard Learning Management System (LMS) webhooks.

Output format

• Executive Summary: Structural Chain Viability (under 150 words)
• Workflow Topology & Node Failure Analysis: (Markdown table containing: Agent Node, Intended Action, Drift Vulnerability, Pedagogical Risk)
• Rubric Fidelity & Bias Diagnostic: (3 numbered sub-sections, 250-350 words total)
• Privacy & Governance Safeguards: (Bullet list detailing data masking and API boundaries)
• Optimization Roadmap: (4-6 prioritized engineering recommendations)

Self-review

1. Have I systematically scrutinized every stage listed in {{grading_chain_nodes}}?
2. Does the rubric analysis explicitly account for the nuances of {{assessment_modality}}?
3. Are data governance points directly tied to {{privacy_compliance_parameters}}?