Skip to content

Latest commit

 

History

History
66 lines (48 loc) · 2.54 KB

analyze-v015.mdx

File metadata and controls

66 lines (48 loc) · 2.54 KB
Raw
title description
Analyzing Results

Interpreting the API Response

The Roundtable Alias API returns a JSON object containing detailed fraud analysis results for each survey response. Key components include:

  • flagged: A boolean indicating if the response is suspected of fraud.
  • num_checks_failed: The number of fraud detection checks the response failed.
  • effort_ratings: A granular assessment of the effort put into each response (1-10 scale).
  • checks: The specific fraud detection checks failed by each response.
{
  "error": false,
  "flagged": true,
  "num_checks_failed": 1,
  "response_groups": {
    "Q1": 1
  },
  "effort_ratings": {
    "Q1": 4
  },
  "checks": {
    "Q1": [
      "Automated test: Off-topic"
    ]
  },
  "model": "alias-v015"
}

Effort Ratings

The effort_ratings object provides an effort score (1-10) for each response, helping you gauge the thoughtfulness and substance of participant answers. Use these ratings to:

  • Quickly identify low-effort responses for review or removal
  • Analyze the distribution of effort across your survey
  • Correlate effort with other response attributes or participant segments

Fraud Detection Checks

The checks object lists the specific fraud signals triggered by each response, such as:

  • Low-effort: The response lacks sufficient detail or substance
  • Automated test: Off-topic: The response is irrelevant to the question asked
  • GPT paste artifacts: The response contains formatting common in AI-generated text
  • Cross-duplicate response: The response is suspiciously similar to another participant's answer

Leverage these granular checks to understand the types of fraud detected and investigate suspicious responses in context.

Customizing Thresholds

Tailor Alias to your specific use case by adjusting the low_effort_threshold parameter. This threshold (1-10) determines the effort rating below which a response is flagged as low quality.

Set a higher threshold to detect only the most egregious low-effort responses, or a lower threshold to cast a wider net. Strike the right balance for your survey goals and tolerance for false positives.

Integration and Next Steps

To start using Alias, simply sign up for an API key and refer to our Get Started guide for step-by-step integration instructions.

For more details on the API request and response schema, see the API Reference page.

If you have any questions or need support, our team is here to help. Reach out at [email protected] or visit our Support Center.