How to use Conversation

Conversation is a special case of multiagent interactions, where only two agents are involved. The configuration file for conversation is similar to the multiagent configuration, but a conversation only has two agents. The following example demonstrates an example:

conversation_head = Conversation('path/to/config.yaml')
interviewer = """Assume you are a researcher and will conduct a human cognition test.
Instruct the subject with the following instructions, one by one. Only ask for one instruction at a time.
Here are the instructions
1) Without further ado, just write the following: "Book."
2) Calculate 4x10.
3) Name three countries with the largest total area in North America.

Start by greeting the subject."""

candidate = (
    "You accepted participating in a research study "
    "and you will follow the instructions of the researchers."
    "Answer with proper sentences."
)

responses = conversation_head.start_conversation(
    intro_prompt_1=interviewer, intro_prompt_2=candidate
)
print(response[0]) # interviewer
print(response[1]) # candidate

Output:

Greetings! Thank you for participating in our human cognition test. Let's begin with the first instruction:
Please write the word "Book."

Book

responses = conversation_head.continue_conversation()
print(response[0]) # interviewer
print(response[1]) # candidate

Output: .. code-block:

Great job! Now, onto the second instruction:
Please calculate 4 multiplied by 10.

4 multiplied by 10 equals 40.