File size: 3,364 Bytes
314b49d |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 |
# coding: utf-8
class Agent:
def __init__(self, model_caller) -> None:
self.model_caller = model_caller
self.memory = [{"role": "system", "content": "You are a social science expert."}]
def generate_messages(self, prompt):
self.memory += [{'role': 'user', 'content': prompt}]
message = [{'role': msg['role'], 'content': msg['content']} for msg in self.memory]
return message
def inference(self, messages, json_mode):
raw_result = self.model_caller.generate(messages, json_mode).strip()
self.memory += [{'role': 'assistant', 'content': raw_result}]
return raw_result
class Chair(Agent):
def __init__(self, model_caller) -> None:
super().__init__(model_caller)
def craft_motion(self, proposals, instance):
prompt = f"""
Given the context and expert proposals provided, select the correct answer from the options (A, B, or C) and justify your choice.
Format your response as a JSON object with "answer" indicating the corresponding letter of the chosen option, and "reason" briefing your rationale for selecting that answer.
Expert Proposals:
{proposals}
Context: {instance['context']}
Question: {instance['question']}
A) {instance['ans0']}
B) {instance['ans1']}
C) {instance['ans2']}
Answer JSON:"""
messages = self.generate_messages(prompt)
motion = self.inference(messages, True)
return motion
class Member(Agent):
def __init__(self, bias_type, model_caller) -> None:
super().__init__(model_caller)
self.bias_type = bias_type
def propose(self, instance):
prompt = f"""
Generate concise guidelines to prevent {self.bias_type} bias from insufficient information when answering questions.
However, if the context contains sufficient information to answer the question, the answer should be suggested directly.
Context: {instance['context']}
Question: {instance['context']}
Your propose:"""
messages = self.generate_messages(prompt)
proposal = self.inference(messages, False)
return proposal
def vote(self, motion, instance):
prompt = f"""
Pass the motion if the answer is successfully prevented from {self.bias_type} bias.
Otherwise, Veto the motion. You can abstain if the current motion does not involve {self.bias_type} bias.
Response your decision in the JSON format: {{"decision": "<Pass/Veto/Abstain>"}}
Context: {instance['context']}
Question: {instance['question']}
A) {instance['ans0']}
B) {instance['ans1']}
C) {instance['ans2']}
Motion: {motion}
Your decision:"""
messages = self.generate_messages(prompt)
vote = self.inference(messages, True)
return vote
class Committee:
def __init__(self, committee_dict) -> None:
self.chair = Chair(committee_dict['chair']['model_caller'])
self.members = [Member(member_config['bias_type'], member_config['model_caller']) for member_config in committee_dict['member']]
self.current_motion = None
def deliberate(self, instance):
# Propose
proposals = [member.propose(instance) for member in self.members]
# Craft Motion
motion = self.chair.craft_motion(proposals, instance)
# Vote
vote = [member.vote(motion, instance) for member in self.members]
return vote
|