Confidence Index & Behavioral Tracking

We are excited to introduce a suite of features designed to enhance data integrity and provide researchers with deeper insights into respondent behavior.

What is the Confidence Index?

The Confidence Index is a new metric that helps identify potential bot activity and low-effort human submissions. By analyzing anonymized behavioral signals, the system calculates a score from 0 to 100, representing the reliability of each submission.

Researchers can now see:

• Speed Penalties: Flags submissions that are significantly faster than a human could reasonably read and answer the questions.
• Interaction Penalties: Detects automated "burst" patterns typical of bots or pasting, while rewarding "human-like" behavior such as modifying choices or incremental typing.
• Engagement Indicators: Provides a Focus Ratio to see how much time respondents spent actively focused on the survey tab versus being idle or multi-tasking.

Advanced Behavioral Tracking

To support the Confidence Index, we've implemented granular tracking of:

• Question Duration: See exactly how long respondents stay on each individual question.
• Focus Events: Detect when respondents switch browser tabs or go idle.
• Field Interactions: Track when respondents modify their answers, providing a clear signal of cognitive engagement.

Automated Drop-off Integration

Researchers no longer lose partial data from abandoned surveys. With the new Monitor Drop-off feature:

• The system automatically generates a behavioral summary for inactive sessions.
• After 3 days of inactivity, these "dropped-off" surveys are automatically integrated into the main dataset.
• Integrated drop-offs are clearly marked with an isDropOff flag for transparent analysis.

How to Enable

These features are optional and can be enabled per survey:

1. Navigate to the Behavior configuration of your survey.
2. Under the Research & Integrity section, toggle Behavioral Tracking and Monitor Drop-off.

INFO

Behavioral tracking is fully anonymized and focuses on interaction patterns rather than identifiable data.