Confidence Scoring: How Polaris Knows What to Trust

Every Polaris brief comes with a confidence score between 0 and 1. This isn't a relevance score — it's an assessment of how trustworthy the information is.

Here's how it works.

What Confidence Measures

It's not about whether the brief is relevant to your query. It's about whether the underlying information is reliable.

The Factors

Confidence is calculated from multiple proprietary signals that we continuously refine. The exact methodology and weighting are proprietary — tuned against real-world accuracy benchmarks across thousands of briefs.

Why This Matters for Agents

Most news APIs return results ranked by relevance. Relevance tells you “this matches your query.” It doesn't tell you “this is true.”

For agents making decisions — trading agents, research agents, policy agents — the distinction is critical. An agent that acts on a 0.45 confidence brief the same way it acts on a 0.92 confidence brief will make bad decisions.

from veroq import PolarisClient

client = PolarisClient()

# Only act on high-confidence intelligence
results = client.search("semiconductor supply chain", min_confidence=0.8)

# Or use confidence for decision weight
for brief in results.briefs:
    if brief.confidence > 0.9:
        # High confidence — act immediately
        process_signal(brief, urgency="high")
    elif brief.confidence > 0.7:
        # Medium confidence — flag for review
        queue_for_review(brief)
    else:
        # Low confidence — monitor only
        add_to_watchlist(brief)

Confidence vs Bias

These are independent scores. A brief can have high confidence (the facts are well-sourced) and high bias (the framing is one-sided). A Reuters report on a political event might be highly confident but still frame the story from a particular angle.

Your agent should check both:

for brief in results.briefs:
    if brief.confidence > 0.8 and brief.bias_score < 0.2:
        # High confidence, low bias — gold standard
        trust_fully(brief)

The Counter-Argument Check

Even high-confidence briefs benefit from the counter-argument. Every brief includes a Devil's Advocate perspective — the strongest case against the brief's main narrative.

This isn't about being contrarian. It's about intellectual honesty. A 0.95 confidence brief about “AI regulation accelerating” still has a counter-argument: “Industry lobbying may delay implementation, and enforcement mechanisms remain undefined.”

Your agent can surface both the brief and its counter-argument to give users a complete picture.

Try It

Every API call includes confidence scores at no extra cost:

curl "https://api.thepolarisreport.com/api/v1/search?q=AI+regulation&min_confidence=0.8&api_key=demo"