# Assessor — Framework overview

Assessor presents the AI competence development framework implemented in the SAICA project. It is a formative diagnostic for self-directed professional development, not a certification or ranking instrument.

## Assessment design

The design uses twelve adaptive questions in a session of approximately twenty minutes. Questions span seven domains, and the selection process prioritizes domain coverage before targeting difficulty. Structured responses and open-ended performance tasks measure both knowledge and applied judgment.

## Seven domains

1. Foundational concepts — understanding models, parameters, and limitations.
2. Prompting — framing requests and improving their results.
3. Workflow design — connecting steps, tools, and human review.
4. Ethics and risk — considering bias, privacy, and responsible use.
5. Tool selection — matching models and approaches to tasks.
6. Output evaluation — checking accuracy and assessing quality.
7. Metacognition — examining assumptions and calibrating confidence.

## Five stages of practice

1. The Copy-Paster — uses outputs with limited checking or adaptation.
2. The Prompt Crafter — prompts deliberately and iterates.
3. The Maker — connects tools and critically evaluates multi-step work.
4. The Engineer — builds reliable workflows with safeguards.
5. The Orchestrator — designs and evaluates AI systems for others.

The personas are shorthand for observable behaviors, not fixed identities or judgments of personal worth.

## Feedback

The intended report includes level placement, a domain profile, explained strengths and growth areas, a question-by-question review, learning resources, and guidance for progressing to the next level.

## Development and validation

Difficulty estimates begin with expert judgment. Pilot calibration, human–AI scoring reliability studies, and fairness analysis are planned. The framework should not be used for hiring, performance reviews, promotion, or credentialing.

This overview summarizes the SAICA repository README, reviewed September 6, 2026. Source project: https://github.com/dliangthinks/saica