Chapter 6 of 17
Types of Prompts
Sorts prompts by their form: declarative, interrogative, instructional and role-based. You see what each one does to the model's answer, how to combine them into a compound prompt, and how to score outputs from different prompt types.
In this chapter
- 6.1 Classifying prompts by sentence form
- 6.2 Declarative prompts
- 6.3 Interrogative prompts
- 6.4 Instructional prompts
- 6.5 Role-based (persona) prompts
- 6.6 Other useful categories
- 6.7 Combining types: the compound prompt
- 6.8 Comparative exercise: one task, three prompt types
- 6.9 Worked numerical
- 6.10 Industry use cases
- 6.11 Review questions
- Knowledge sources
6.1 Classifying prompts by sentence form
In grammar, sentences are declarative (statements), interrogative (questions) or imperative (commands). The syllabus uses the same idea for prompts.
| Type | Form | What it does to the model | Best for |
|---|---|---|---|
| Declarative | A statement of fact or situation | Sets context or a starting point for the model to continue or build on | Giving background, setting a scene, starting a text the model completes |
| Interrogative | A question | Asks the model to retrieve, explain or reason toward an answer | Explanations, facts, opinions, analysis |
| Instructional | A command with an action verb | Tells the model exactly what to produce and how | Producing a defined output: emails, tables, code, summaries |
6.2 Declarative prompts
A declarative prompt states something and lets the model take it forward. On its own it is open-ended, so the model decides what to do with it.
Examples
- "Our company sells organic tea in tier-2 cities and our online sales fell 15% last quarter."
- "The following is a formal notice from the principal to all parents regarding the annual sports day." (The model continues by writing the notice.)
Strengths: good for setting context; natural way to start a piece of writing.
Weaknesses: the model has to guess the task. Followed by nothing else, the first example above might get an analysis, a list of causes, or a sympathetic comment.
Best use: as the context part of a larger prompt, followed by a question or instruction.
6.3 Interrogative prompts
An interrogative prompt asks a question. The model answers it.
Types of questions
| Question type | Example | Output |
|---|---|---|
| Closed (factual) | "What is the GST rate on restaurant services in India?" | Short, specific answer (must be verified) |
| Open (explanatory) | "Why do small businesses struggle with cash flow?" | Explanation with reasons |
| Analytical | "What are the advantages and disadvantages of remote work for an IT company?" | Structured comparison |
| Hypothetical | "What might happen to a dairy cooperative's income if milk procurement prices rise by 10%?" | Reasoned scenario |
| Socratic (for learning) | "Ask me questions one at a time to test my understanding of photosynthesis." | Interactive tutoring |
Strengths: natural to write; good for learning and exploring.
Weaknesses: the output's length, depth and format are left to the model unless you specify them.
6.4 Instructional prompts
An instructional prompt gives a direct command, usually starting with an action verb, and often states the format and limits.
Examples
- "Summarise the attached report in five bullet points for a busy manager."
- "Convert this paragraph into a table with columns: Product, Price, Stock."
- "Write a Python function that checks whether a number is prime, with comments."
- "Translate this notice into Hindi, keeping a formal tone."
Step-wise instructions. For multi-part tasks, number the steps:
1. Read the customer complaint below.
2. Identify the main issue in one sentence.
3. Classify it as Billing, Delivery, Product Quality or Other.
4. Draft a reply under 100 words.
Complaint: """..."""
Strengths: the most control over the output; repeatable results; easy to turn into templates.
Weaknesses: a rigid instruction may stop the model from offering a better idea you had not thought of.
6.5 Role-based (persona) prompts
A role prompt asks the model to take on a role, which shapes vocabulary, depth and tone.
| Role | Effect |
|---|---|
| "You are a chartered accountant." | Uses tax and accounting terms; careful, compliance-minded tone |
| "You are a friendly primary school teacher." | Simple words, short sentences, encouraging tone |
| "You are a strict peer reviewer for a journal." | Critical, finds weaknesses |
| "You are an interviewer for a software developer job." | Asks questions one by one, gives feedback |
A role does not give the model expertise it lacks; it steers style and focus. Facts still need checking.
6.6 Other useful categories
- Comparative prompts: "Compare UPI and credit cards for small merchants on cost, speed and risk, in a table."
- Conditional prompts: "If the text mentions a refund, extract the amount; otherwise write 'No refund'."
- Creative prompts: "Write a four-line jingle for a Kanpur leather goods brand."
- Reflective prompts: "Review your answer above. List any claims that might be wrong."
6.7 Combining types: the compound prompt
The strongest prompts usually combine all three forms: a declarative part for context, an interrogative part for the problem and an instructional part for the output.
[Declarative] Our hospital's outpatient waiting time averages 95 minutes, and patient complaints have doubled this year.
[Interrogative] What are the most likely causes, and which can be fixed within three months?
[Instructional] Answer as a hospital operations consultant. Give a table with Cause, Evidence to check, Quick fix, and Cost (Low/Medium/High). Limit to six rows.
Microsoft's guidance for Copilot makes a similar point. It suggests every prompt state a goal, context, expectations (format and tone) and, where useful, a source of information to draw on.
6.8 Comparative exercise: one task, three prompt types
Task: produce material on the benefits of drip irrigation for a farmer awareness leaflet.
| Prompt type | Prompt | Typical outcome |
|---|---|---|
| Declarative | "Drip irrigation saves water." | The model expands the statement in a general way; may not suit a leaflet. |
| Interrogative | "What are the benefits of drip irrigation for small farmers in Uttar Pradesh?" | A good explanation, but length and style vary. |
| Instructional | "Write a leaflet section in simple Hindi for small farmers in Uttar Pradesh on drip irrigation. Give four benefits as bullets, one line on government subsidy schemes to ask about at the block office, and a call to action. Under 120 words." | Ready-to-use content, close to the final need. |
6.9 Worked numerical
Numerical 6.1: Scoring outputs from different prompt types.
Five evaluators rate the drip irrigation outputs from Section 6.8 on a 1–5 scale for fitness for use.
| Evaluator | Declarative | Interrogative | Instructional |
|---|---|---|---|
| E1 | 3 | 3 | 5 |
| E2 | 2 | 4 | 4 |
| E3 | 3 | 3 | 5 |
| E4 | 3 | 3 | 4 |
| E5 | 2 | 4 | 5 |
Mean score:
Sample standard deviation for each column:
For the declarative column the deviations are 0.4, −0.6, 0.4, 0.4, −0.6. Their squares sum to 0.16 + 0.36 + 0.16 + 0.16 + 0.36 = 1.20, so s = √(1.20 / 4) = √0.30 ≈ 0.548. By the same working, the other two columns also give s ≈ 0.548.
The evaluators agreed equally well on all three, but the instructional prompt scored 2 points higher on average than the declarative one. With only five raters this is a classroom illustration, not a formal test; a larger study would use more raters and a significance test.
6.10 Industry use cases
- Customer support ticket triage. Support teams commonly use instructional, conditional prompts to classify incoming tickets by category and urgency and to extract order numbers, before a human or a workflow routes them.
- Interview practice. Recruitment and training teams use role-based prompts ("You are an interviewer for a bank PO post; ask me one question at a time") so candidates can practise.
- Workspace assistants. Microsoft's Copilot guidance (goal, context, expectations, source) and Google's Gemini guidance (persona, task, context, format) both teach users to combine context-setting statements with clear instructions.
6.11 Review questions
- Define declarative, interrogative and instructional prompts with two examples each.
- Why is a declarative prompt alone often a weak prompt? How can it be strengthened?
- Write a compound prompt for analysing why a college's placement rate has fallen.
- Give four types of interrogative prompts with an example from banking.
- Explain what a role prompt can and cannot do.
- Compute the mean and sample standard deviation for ratings 4, 5, 3, 4, 4. (Answer: mean 4.0, s ≈ 0.707.)
Knowledge sources
- Microsoft Support, Learn about Copilot prompts.
- Google, Gemini for Google Workspace: Prompting Guide 101, 2024.
- OpenAI, Prompt engineering guide.
- Schulhoff, S. et al., The Prompt Report: A Systematic Survey of Prompting Techniques, 2024.
Independent study notes for learning purposes. Not affiliated with any university or course provider. Facts and case-study figures are linked to their original sources. ChatGPT, Gemini, Copilot, Jasper, Synthesia, Runway, Lumen5 and other product names are trademarks of their respective owners.