2025-11-25 19:28:26 +01:00

4.9 KiB

Raw Blame History

sidebar_position
5

Opinions

How agents form, store, and evolve beliefs.

import Tabs from '@theme/Tabs'; import TabItem from '@theme/TabItem';

What Are Opinions?

Opinions are beliefs formed by the agent based on evidence and personality. Unlike world facts (objective information received) or agent facts (actions taken), opinions are judgments with confidence scores.

Type	Example	Confidence
World Fact	"Python was created in 1991"	—
Agent Fact	"I recommended Python to Bob"	—
Opinion	"Python is the best language for data science"	0.85

How Opinions Form

Opinions are created during think operations when the agent:

Retrieves relevant facts
Applies personality traits
Forms a judgment
Assigns a confidence score

graph LR
    F[Facts] --> P[Personality Filter]
    P --> J[Judgment]
    J --> O[Opinion + Confidence]
    O --> S[(Store)]

# Ask a question that might form an opinion
answer = client.think(
    agent_id="my-agent",
    query="What do you think about functional programming?"
)

# Check if new opinions were formed
for opinion in answer["new_opinions"]:
    print(f"New opinion: {opinion['text']}")
    print(f"Confidence: {opinion['confidence']}")

Searching Opinions

# Search only opinions
opinions = client.search_memories(
    agent_id="my-agent",
    query="programming languages",
    fact_type=["opinion"]
)

for op in opinions:
    print(f"{op['text']} (confidence: {op['confidence_score']:.2f})")

hindsight memory search my-agent "programming" --fact-type opinion

Opinion Evolution

Opinions change as new evidence arrives:

Evidence Type	Effect
Reinforcing	Confidence increases (+0.1)
Weakening	Confidence decreases (-0.15)
Contradicting	Opinion revised, confidence reset

Example evolution:

t=0: "Python is best for data science" (0.70)
     ↓ New evidence: Python dominates ML libraries
t=1: "Python is best for data science" (0.85)
     ↓ New evidence: Julia is 10x faster for numerical computing
t=2: "Python is best for data science, though Julia is faster" (0.75)
     ↓ New evidence: Most teams still use Python
t=3: "Python is best for data science" (0.82)

Personality Influence

Different personalities form different opinions from the same facts:

# Create two agents with different personalities
client.create_agent(
    agent_id="open-minded",
    personality={"openness": 0.9, "conscientiousness": 0.3, "bias_strength": 0.7}
)

client.create_agent(
    agent_id="conservative",
    personality={"openness": 0.2, "conscientiousness": 0.9, "bias_strength": 0.7}
)

# Store the same facts to both
facts = [
    "Rust has better memory safety than C++",
    "C++ has a larger ecosystem and more libraries",
    "Rust compile times are longer than C++"
]
for fact in facts:
    client.store(agent_id="open-minded", content=fact)
    client.store(agent_id="conservative", content=fact)

# Ask both the same question
q = "Should we rewrite our C++ codebase in Rust?"

answer1 = client.think(agent_id="open-minded", query=q)
# Likely: "Yes, Rust's safety benefits outweigh migration costs"

answer2 = client.think(agent_id="conservative", query=q)
# Likely: "No, C++'s ecosystem and our team's expertise make it the safer choice"

Bias Strength

The bias_strength parameter (0-1) controls how much personality influences opinions:

Value	Behavior
0.0	Pure evidence-based reasoning
0.5	Balanced personality + evidence
1.0	Strongly personality-driven

# Evidence-focused agent
client.create_agent(
    agent_id="analyst",
    personality={"bias_strength": 0.2}  # Low bias
)

# Personality-driven agent
client.create_agent(
    agent_id="advisor",
    personality={"bias_strength": 0.8}  # High bias
)

Opinions in Think Responses

When think uses opinions, they appear in based_on:

answer = client.think(agent_id="my-agent", query="What language should I learn?")

print("World facts used:")
for f in answer["based_on"]["world"]:
    print(f"  {f['text']}")

print("\nOpinions used:")
for o in answer["based_on"]["opinion"]:
    print(f"  {o['text']} (confidence: {o['confidence_score']})")

Confidence Thresholds

Opinions below a confidence threshold may be:

Excluded from responses
Marked as uncertain
Revised more easily

# Low confidence opinions are held loosely
# "I think Python might be good for this" (0.45)

# High confidence opinions are stated firmly
# "Python is definitely the right choice" (0.92)

4.9 KiB Raw Blame History