File system
# TruthBot: Parallel Truthful Inference for Detecting and Mitigating AI Manipulation
**Julius Chandler**
REIUK LTD; University of California, Berkeley
*With Apart Research*
---
## Abstract
Large language models can be prompted to provide strategically biased or manipulative responses that selectively present evidence to influence user beliefs. We present TruthBot, a dual-purpose framework that both mitigates AI manipulation and detects manipulative intent. Our approach deploys a guaranteed-truthful parallel model alongside any AI system, providing users with comprehensive, balanced information that neutralizes selective
On the state of this archive
Over the years I have not documented my writing (as much as I have apropos my art, or my signalling). Therefore the current selection is limited to what I have on my laptop and one of my newer drives. Additionally, I believe the season for me to write is not yet ripe. It is more wise at my age to listen (or read!) and in later years write.