Prompt Chaining is a technique where multiple prompts are linked together to generate a sequence of outputs from the model.
The idea is that by breaking prompts down into smaller, manageable sub-tasks (handled by a separate prompt), the model can perform more complex tasks.
Here’s a detailed breakdown of different types of prompt chaining:
Serial Prompting
This involves providing the model with a sequence of prompts, one after the other, where each prompt builds upon the previous one.
Prompt 1
List the necessary elements of a contract in UK Contract Law. Provide no extra details, just the elements in a list.
Output 1
1. Offer
2. Acceptance
3. Consideration
4. Intention to create legal relations
5. Capacity to contract
6. Certainty of terms
Then in the same chat window, follow up with this prompt:
Prompt 2
Provide a brief explanation for each of the above.
Output 2
Offer: A clear and definite proposal made by one party (the offeror) to another (the offeree) to form a contract, with the intention that it will become binding upon acceptance.
Acceptance: An unequivocal agreement to the terms of the offer by the offeree. Acceptance must be communicated to the offeror and match the terms of the offer exactly.
Consideration: Something of value exchanged between the parties, which each party gives as the price of the other's promise. This can be a benefit to one party or a detriment to the other.
Intention to create legal relations: Both parties must intend for the agreement to be legally binding and enforceable by law.
Capacity to contract: The parties involved must have the legal ability to enter into a contract, which generally means they must be of sound mind, not under duress, and of legal age.
Certainty of terms: The terms of the contract must be clear and certain enough for the courts to enforce. This includes clarity on the subject matter and the obligations of each party.
Hierarchical Prompting
Hierarchical prompting uses a high-level prompt that is broken down into lower-level prompts, each refined until the task is completed.
High-Level Prompt 1
Write a mock legal brief for a breach of contract case. Use UK Law.
{{You can also attach a breach of contract case if you would like to focus on
a real case}}
Low-Level Prompt 1
Outline the relevant laws and precedents.
Low-Level Prompt 2
Analyse the application of these laws to the facts of the case
Low-Level Prompt 3
Conclude with the likely outcome based on the analysis
Hybrid Prompting
This approach combines both serial prompting and hierarchical prompting to handle more complex tasks.
High Level Prompt 1
Generate a comprehensive legal report on data protection compliance
Serial Prompt 1
List the key components required in a data protection compliance report
Output 1
Introduction, Legal Framework, Data Collection Practices, Data Security Measures, Compliance Recommendations.
Hierarchial Prompt 1
Expand on each component.
Low-Level Prompt 1
Detail data collection practices and their compliance with legal standards.
Low-Level Prompt 2
Provide compliance recommendations based on current practices and legal requirements.
Steps to Effective Prompt Chaining
By chaining prompts together effectively, ChatGPT can generate longer sequences of text or perform more complex tasks with greater accuracy and coherence.
Define the Main Task: Clearly state the overall task you want the LLM to accomplish.
Break Down the Task: Divide the main task into smaller, logical sub-tasks.
Sequential Prompts: Create a sequence of prompts where each prompt builds on the previous one.
Hierarchical Prompts: Develop a structure where high-level prompts are broken down into more specific prompts.
Combine Techniques: Use a combination of serial and hierarchical prompts for complex tasks.
By following these steps and using well-structured examples, legal professionals can leverage prompt chaining to handle complex legal tasks more efficiently and accurately.
