Research Notes

How Bouncer detects influence, where it fails, and what we're doing about it. See also: our live methodology

2026-03-15

Is the Model Politically Biased? A 120-Video Calibration Study

We analyzed 12 channels across conservative, progressive, and independent orientations. The model wasn't biased — but it was consistently wrong about transparency for all partisan commentary. Here's the data and the fix.

2026-03-15

The Psychology of Podcast Influence: Why Long-Form Audio Bypasses Your Defenses

Podcast listeners process influence differently than video viewers. The 2-hour conversation format, parasocial trust, and conversational consensus create unique vulnerability patterns that standard detection models miss.

2026-03-09

The Science of Influence: Academic Foundations of Automated Detection

A comprehensive mapping of Bouncer's six influence dimensions and 29 technique definitions to their foundations in social psychology, rhetoric, and media studies.

2026-03-09

Inoculation at Scale: From Prebunking to Real-Time Influence Transparency

How McGuire's inoculation theory, van der Linden's prebunking research, and computational propaganda detection converge in a real-time transparency system.

2026-03-09

Story Shaping: How Stories Replace Arguments

Our #1 detected influence dimension. When story shaping scores high, transparency drops by half.

2026-03-08

Cross-Spectrum Calibration: Detecting Influence Regardless of Ideology

Our analysis was better at detecting conservative-coded manipulation than progressive-coded manipulation. Here's the data, the fix, and the results.