AIT-LIT-01
Generative AI explanation
Explains model inputs, outputs, uncertainty, and limitations accurately enough to guide professional use.
Loading your workspace…
Loading your workspace…
AI & TECHNOLOGY · COMPLETE DRAFT · READY FOR REVIEW
A source-grounded course for understanding how generative AI works, what its outputs can and cannot establish, and how to use it with verification, privacy, and accountable human judgment.
GOVERNANCE GATE
This version exposes the proposed sources, competencies, instruction, evidence requirements, and rubric for accountable review. It cannot be enrolled in, completed, or used to issue a credential until an authorized reviewer approves a final immutable version and the program is separately published.
Status: Complete draft awaiting accountable instructional review
Draft version: 1.0.0-draft.1
Estimated learning time: 150 minutes
Program level: Course
Evidence activities: 3
Credential pathway: Proposed Certificate of Completion after review, publication, learner evidence approval, and separate administrator issuance.
LEARN → APPLY → DEMONSTRATE → REVIEW → VERIFY
AIT-LIT-01
Explains model inputs, outputs, uncertainty, and limitations accurately enough to guide professional use.
AIT-LIT-02
Evaluates claims, sources, and failure patterns through proportionate independent checks.
AIT-LIT-03
Defines privacy, disclosure, authorship, fairness, and human-decision boundaries for a generative-AI workflow.
CONTINUING PROFESSIONAL CASE
A team uses generative AI to draft executive briefings from public reports and internal notes. The drafts are fluent, but one invented a source and another exposed sensitive project details in a third-party service.
Constraint: The team needs a usable workflow next month and cannot assume that every employee has technical expertise.
Learner task: Design a literacy-based workflow that defines appropriate use, verification, privacy, disclosure, and accountable approval.
FIVE SOURCE-GROUNDED LESSONS
LESSON 1
Purpose: Explain generated output without mistaking fluency for grounded knowledge.
A generative model predicts and constructs outputs from learned patterns and current context. Fluency can coexist with factual error, missing provenance, or unstable reasoning.
Professional example: A polished market summary contains a fabricated statistic because the model optimized a plausible continuation rather than checking a source.
Evidence connection: NIST identifies confabulation and information-integrity risks as characteristic generative-AI concerns.
Label each claim in the briefing as generated, source-supported, uncertain, or requiring subject-matter review.
Which statement is most accurate?
Expected: Generative output can be useful while still requiring source and context checks. — Usefulness and reliability are separate questions.
LESSON 2
Purpose: Choose bounded uses based on stakes and evidence.
The right question is not whether AI can produce an output, but whether that output supports a defined purpose under acceptable error, privacy, and review conditions.
Professional example: Drafting headings is lower risk than deciding which employee should receive discipline.
Evidence connection: NIST risk management begins with context, intended purpose, impacts, and risk tolerance.
Separate drafting assistance from factual approval and executive sign-off in the briefing workflow.
What best establishes task fit?
Expected: The purpose, stakes, evidence, and correction process are explicit. — Task fit is contextual and evidence-dependent.
LESSON 3
Purpose: Use prompting as interface design rather than incantation.
Clear instructions can specify audience, inputs, boundaries, structure, and uncertainty behavior, but they cannot guarantee truth or remove the need for evaluation.
Professional example: A briefing prompt requires claim-by-claim source references and an explicit 'insufficient evidence' response.
Evidence connection: Current API guidance treats instructions, structured outputs, and evaluation as connected design choices.
Write an output contract that separates sourced facts, analysis, and unresolved questions.
What is the strongest prompting practice?
Expected: Define inputs, output structure, evidence expectations, and uncertainty behavior. — An output contract makes behavior more testable.
LESSON 4
Purpose: Check consequential claims using evidence independent of the output.
Verification traces a claim to an authoritative source, checks whether the source supports it, and records unresolved uncertainty. Repeating the question to the same model is not independent confirmation.
Professional example: The analyst opens the cited report, confirms the statistic and date, and records the source location.
Evidence connection: NIST emphasizes content provenance, information integrity, testing, and monitoring.
Create a verification checklist for every factual claim included in the executive briefing.
Which action provides the strongest verification?
Expected: Check the claim against the cited authoritative source and its context. — Verification requires evidence independent of generated fluency.
LESSON 5
Purpose: Make responsible-use boundaries operational.
A professional workflow should define permitted data, material AI disclosure, who approves consequential outputs, how errors are corrected, and when use must stop.
Professional example: Internal notes are excluded from the public tool, drafts are labeled, and a named analyst approves sourced claims.
Evidence connection: UNESCO, OECD, and NIST converge on privacy, transparency, accountability, and human-centered governance.
Publish a one-page operating policy for the briefing assistant, including correction and stopping rules.
What makes human oversight meaningful?
Expected: A named person can inspect evidence, reject the output, and stop the workflow before use. — Oversight needs evidence, authority, and timely intervention.
ASSESSED APPLICATION
PROPOSED REVIEW RUBRIC
Problem framing and operational boundary is explicit, workable, and supported by relevant evidence.
Competency: AIT-LIT-01
Technical design and implementation rationale is explicit, workable, and supported by relevant evidence.
Competency: AIT-LIT-02
Evaluation design and preserved evidence is explicit, workable, and supported by relevant evidence.
Competency: AIT-LIT-03
Risk controls and accountable governance is explicit, workable, and supported by relevant evidence.
Competency: AIT-LIT-03
Failure analysis, revision, and professional judgment is explicit, workable, and supported by relevant evidence.
Competency: AIT-LIT-03
PRIMARY SOURCE BASE
National Institute of Standards and Technology · NIST AI 600-1 · 2024
Provides a cross-sector framework for identifying, measuring, and managing generative-AI risks across the lifecycle.
UNESCO · UNESCO · 2023
Provides a human-centered account of generative-AI capabilities, data privacy, agency, validation, and responsible use.
Organisation for Economic Co-operation and Development · OECD.AI · Updated 2024
Provides internationally recognized principles for trustworthy AI, transparency, robustness, accountability, and human-centered values.