Tips for Effective Prompts
The quality of PLAI responses depends significantly on well-formulated prompts. A well-crafted prompt is the key to precise, informative, and relevant answers that are clear and logical. The language used by the language model is also crucial: Do you want to create a legal opinion that is understandable to laypeople, or is it intended for a legal professional?
1. Precision is Crucial
Formulate your prompt as clearly and precisely as possible. Avoid vague or ambiguous language to ensure the model correctly interprets your intentions.
Example:
Poor: “Create an employment contract for me.”
This prompt is too general and does not provide enough information about what should be included in the employment contract. It lacks precision, leading to a generic or incomplete response.
Good:
“Create a legally binding employment contract for a full-time position. Include: employer’s name, employee’s name, start date, working hours, compensation, vacation entitlement, and termination procedures. Consider all relevant legal provisions and ensure clear wording.”
This prompt is more precise as it gives specific instructions on the required contents of the employment contract. By specifying key elements such as names, working hours, and legal aspects, it ensures that the generated response is detailed and meets the requirements.
2. Provide Context
Supply enough context to help the model understand the desired focus. The clearer the context, the more precise the model’s responses will be. Provide background information or specific requirements to assist the model in better understanding the given task. Good context offers clear instructions, relevant details, and specific conditions.
Poor Context Prompt:
“Create an employment contract for a graphic designer.”
This prompt provides little context about specific requirements or details to be considered in the employment contract. It could lead to a general and potentially inappropriate response.
Good Context Prompt:
“Create an employment contract for a creative design position in a graphic design company. The contract should include specific tasks and responsibilities, working hours, flexible work arrangements, intellectual property, and data protection provisions. Ensure the contract reflects the particular requirements of the design industry and complies with current legal regulations.”
This prompt provides context about the specific position (creative design), the industry (graphic design), and certain aspects to be included in the contract. The context helps the model generate more relevant and tailored information.
3. Structure Complex Requests
For complex requests, it is helpful to provide a structure and an organizational framework. This assists the model in understanding and generating coherent information.
Example:
Poor: “Create a contract for renting a residential space. It should cover things like rent, deposit, and duties of the landlord and tenant.”
Good: “Draft a legal contract for renting a residential space. Organize the contract into the following sections:
1. Parties to the Contract: Provide names and addresses of the landlord and tenant.
2. Lease Terms: Clearly specify monthly rent, deposit, and payment terms.
3. Landlord’s Duties: List the responsibilities of the landlord, including maintenance obligations.
4. Tenant’s Duties: Clearly define the tenant’s obligations, including compliance with house rules.
5. Duration of the Contract: Establish the start and end dates of the lease.
6. Ensure that the contract complies with local rental laws.”
This example is well-structured, using clear sections and bullet points. The structure makes it easier for the model to understand the requirements and generate precise information. If you have an idea of the outline or structure you desire, communicate it.
4. Specify the Format
If you want a specific format for the output, make it clear—whether it’s bullet points (1., 2., 3… or A., B., C., or further subdivisions according to a specific format, such as A, I, 1, a, aa, aaa). Also, specify paragraphs, indentations, and the number of line breaks after certain points precisely.
5. Avoid Ambiguities
It is crucial to avoid ambiguities in your prompts to prevent misunderstandings.
Poor: “Describe the employer’s duties regarding overtime regulations.”
This example is vague and could lead to different interpretations. Which specific aspects of overtime regulations should be addressed? This could result in an unclear or incomplete response or at least not the intended answer.
Good: “Clarify the legal responsibilities of the employer regarding overtime regulations. Consider set limits for overtime, the need for written agreements, and compensation for additional work hours. Ensure the response references current labor laws.”
Explanation: This example is more precise, addressing specific aspects relevant to overtime regulations. It minimizes the possibility of misunderstandings and helps the model provide accurate and relevant information.
6. Clarify What is Not Desired
In addition to specifying the desired requirements in your prompts, it is equally important to explain what you do not want in the response. This can pertain to specific content aspects (e.g., “do not examine deception”), certain formulations (“avoid words like…,” “do not use phrases like…”), or even formatting preferences (e.g., “do not use bullet points; write arguments in continuous text”).
7. Trial and Error: Experiment
Experiment with different formulations to achieve the best results. Sometimes the model is sensitive to specific word choices, so it’s worth trying various approaches.
