Artificial intelligence against misinformation among students | Miloš Hrůza

The project focuses on testing current AI tools in the fight against misinformation among university students. The aim was to test the accuracy and reliability of the fact-checking capabilities of popular chatbots and compare them with verified sources such as Demagog.cz, FactCheck.org, and Google Fact Check Explorer.

The research gradually evolved from a general concept to practical testing of four specific claims using ChatGPT, Google Gemini, and Perplexity.ai. The tested claims included political disinformation (Trump vs. Biden), technical facts (aviation safety), local Czech issues (Prymula and COVID-19), and general knowledge (bananas as botanical berries). The project also tested the influence of prompt language on the quality of responses.

It included an analysis of a dataset of fake news from Kaggle, the creation of a prototype fact-checking chatbot using Character.ai, and the design of a user interface. Special attention was paid to data visualization and the creative presentation of results.

Ethics: The project comprehensively addressed ethical aspects using Tarot Cards of Tech and Ethics Canvas. It identified risks such as bias in training data, false certainty of AI, privacy violations, and possible misuse for spreading misinformation. Proposed measures include warning about the uncertainty of outputs, using only validated sources, ensuring algorithm transparency, and testing model boundaries.

AI TOOLS USED IN THE PROJECT

1. ChatGPT (free/paid) – main tool for brainstorming, project formulation, fact-checking testing, data analysis, visualization generation
2. Claude – comparison of outputs with ChatGPT, alternative perspectives, methodology suggestions
3. Google Gemini – fact-checking testing, comparison with other chatbots
4. Perplexity.ai – fact-checking testing, analysis of information sourcing
5. Character.ai – creation of a fact-checking chatbot prototype
6. DALL-E – storyboard generation for educational purposes (with limitations when working with text)
7. FigJam AI (Figma) – creation of a wireframe for a fact-checking application
8. Gamma AI – automatic creation of a project presentation outline
9. Kaggle – source of a dataset of fake news for analysis and visualization

 LINKS TO OUTCOMES

• Project presentation
• Dataset for analysis

Note: Some outputs are integrated directly into the final document and presentation. The chatbot prototype had limited access to the internet, which restricted its practical use.