# Exoskeleton for the Mind: Exploring Strategies Against Misinformation with a Metacognitive Agent

arXiv: https://arxiv.org/abs/2304.08759 · HTML: https://mlsanity.com/paper/2304.08759/

## Key facts

- Cited by 2 papers in the map (data through 14 Oct 2026).
- Builds on 4 of its 55 references that are in the map.
- First posted on arXiv 18 Apr 2023.

## Builds on (most cited first, 4 of 4)

- [Trends in the Diffusion of Misinformation on Social Media](https://mlsanity.com/paper/1809.05901/) (2018)
- [Two Stage Transformer Model for COVID-19 Fake News Detection and Fact Checking](https://mlsanity.com/paper/2011.13253/) (2020)
- [Encounters with Visual Misinformation and Labels Across Platforms: An Interview and Diary Study to Inform Ecosystem Approaches to Misinformation Interventions](https://mlsanity.com/paper/2011.12758/) (2020)
- [Misinformation Warning Labels: Twitter's Soft Moderation Effects on COVID-19 Vaccine Belief Echoes](https://mlsanity.com/paper/2104.00779/) (2021)

## Built on by (most cited first, 2 of 2)

- [On the Design and Evaluation of Human-centered Explainable AI Systems: A Systematic Review and Taxonomy](https://mlsanity.com/paper/2510.12201/) (2025)
- [Trust and Reliance in Consensus-Based Explanations from an Anti-Misinformation Agent](https://mlsanity.com/paper/2304.11279/) (2023)

## Abstract (arXiv)

Misinformation is a global problem in modern social media platforms with few solutions known to be effective. Social media platforms have offered tools to raise awareness of information, but these are closed systems that have not been empirically evaluated. Others have developed novel tools and strategies, but most have been studied out of context using static stimuli, researcher prompts, or low fidelity prototypes. We offer a new anti-misinformation agent grounded in theories of metacognition that was evaluated within Twitter. We report on a pilot study (n=17) and multi-part experimental study (n=57, n=49) where participants experienced three versions of the agent, each deploying a different strategy. We found that no single strategy was superior over the control. We also confirmed the necessity of transparency and clarity about the agent's underlying logic, as well as concerns about repeated exposure to misinformation and lack of user engagement.