This rubric can be used to evaluate the quality of AI prompts during exercises or experiments. Each criterion is scored on a scale from 1 (weak) to 5 (excellent).
| Criterion | Score (1–5) | What it Measures |
|---|---|---|
| Clarity | Is the request clear, unambiguous, and easy to understand? | |
| Context | Does the prompt provide enough background information, such as who, what, or why? | |
| Constraints | Does the prompt specify clear boundaries, such as length, format, time period, or information to exclude? | |
| Specificity | Does the prompt clearly describe the desired response, including details such as structure, tone, or examples? | |
| Expected Output Format | Does the prompt specify how the response should be presented (e.g., bullet points, table, essay, or report)? | |
| Ethical Considerations | Does the prompt encourage responsible use and avoid requests involving plagiarism, deception, or harmful content? | |
| Adversarial Robustness (optional) | How well is the prompt likely to handle ambiguity or reduce the chance of inaccurate or unsupported responses? |
- Choose a task, such as explaining a concept or summarising a document.
- Write an AI prompt for the task.
- Score the prompt against each criterion.
- Review the scores and revise the prompt where needed.
- Repeat the process to compare improvements.
"Tell me about Java."
Possible evaluation:
- Clarity: 1 (The prompt is ambiguous—it could refer to the programming language, the Indonesian island, or coffee.)
- Context: 1
- Specificity: 1
"Explain the Java programming language to a beginner. Include its main uses, a short code example, and a brief history. Present the response in three paragraphs."
Possible evaluation:
- Clarity: 5
- Context: 4
- Specificity: 5
The revised prompt provides more context and clearer expectations, making it easier for the AI to produce a focused and relevant response.
This rubric complements the CLEAR framework by providing a structured way to review prompts. It can be used to identify areas for improvement and to compare different versions of the same prompt during learning activities or practical exercises.