NewsBattery – User-Controlled AI Political News Adapatation How does a system providing user-controller features like language intensity reduction and summarization of political news articles affect users’ news reading behavior?

Abstract

News overload, avoidance, and fatigue have become significant concerns in news consumption in recent years. In an increasingly fast-changing world and media environment, where misinformation spreads easily, journalism plays a critical role. However, an increasing amount of people disengage from news for reasons ranging from disinterest to protection of their mental health. While content-focused strategies such as reporting more good news and providing perspectives in articles can help, this thesis explores a technical solution to address issues like news overload, avoidance, and fatigue. Recent advances in artificial intelligence, particularly Large Language Models (LLMs) have shown remarkable capabilities in text generation. In the context of this thesis NewsBattery is developed. It offers user-controlled rewriting of political articles via language intensity reduction, summarization and casualty number replacement with general categories. This thesis investigates how prompt engineering can be used to reduce the language intensity of political news articles while maintaining factual accuracy. Results reveal that broadly framed, minimally restrictive prompts outperform highly prescriptive or persona-driven prompts in balancing reduced language intensity and factuality. Additionally, this thesis evaluates the effects of the user-controlled news adaptation system NewsBattery on news overload, avoidance, fatigue, analysis paralysis, need for news, trust in news and perceived political information efficacy (PPIE). A comparison of participants’ responses before and after using NewsBattery indicates significant reductions in news fatigue and analysis paralysis, along with significant improvements in PPIE. Furthermore, trust in NewsBattery is higher than general trust in news, and usability ratings are excellent. By reducing emotional burden while preserving core content, NewsBattery is an example of how AI can be used to support healthier and more accessible engagement with political news consumption.

Date of Award	2025
Original language	English
Supervisor	Andreas Stöckl (Supervisor)