🚧 This site is under development; its content is not final and may change at any time. 🚧
Basic Prompt EngineeringAssign the model a job

Assign the model a job

Introduction to Role Prompting

Imagine going to the doctor with a mysterious ache, but instead of giving details, you just say, “Doc, something feels wrong.” The doctor, scratching their head, tries guessing: “Stomachache? Headache? Heartbreak?” Frustration builds as both of you stumble through a guessing game, wasting time and missing the mark.

In role prompting, you assign the model a defined role or profession, such as “teaching assistant,” “market researcher,” or “business strategist.” This role acts as a framework, guiding the AI’s tone, language, and depth of response. Instead of a generic answer, the model aligns its output with the expectations and responsibilities of the assigned role.1

Benefits of Role Prompting

  1. Focus and Expertise: Assigning a role encourages the model to draw on relevant knowledge and expertise. This approach yields more nuanced and specialized responses.
  2. Contextual Language: Role prompts shape the tone and language of responses to fit the audience. For example, a “legal advisor” role ensures formal, precise language, while a “personal assistant” role prioritizes a friendly, conversational tone.
  3. Creativity and Structure: For tasks requiring creativity or complex outputs, role prompting adds structure. A model acting as a “storyteller,” for instance, will follow narrative conventions rather than providing purely factual content.

By giving the AI a clear role and objective (much like telling the doctor exactly where it hurts), you eliminate guesswork, streamline interactions, and enhance the quality of AI-generated responses.

Examples of Role Prompting

🙋
Analyze this Data. […]
🤖
[Unstructured and unhelpful response]

🙋

You are a financial analyst explaining trends to a non-expert audience. Analyze this data. […]

🤖
[Clear, audience-tailored response]

References & Footnotes

Footnotes

  1. Kong, A., Zhao, S., Chen, H., Li, Q., Qin, Y., Sun, R., Zhou, X., Wang, E., & Dong, X. (2023). Better zero-shot reasoning with role-play prompting. arXiv. https://doi.org/10.48550/ARXIV.2308.07702 ↩

Last updated on