With the advancement of information technology, social media platforms have become a part of everyday life. Consequently, manipulating information has posed challenges for many democracies. Amidst geopolitical tensions, Taiwan—which has emerged from a history of authoritarianism to embrace democracy and freedom—has become a major target for information manipulation by authoritarian regimes. According to Sweden’s V-Dem Institute, for instance, Taiwan is ranked as the country most severely affected by information manipulation worldwide for many years.
In 2022, after OpenAI launched ChatGPT, generative artificial intelligence (GenAI) redefined work methods and paradigms by efficiently producing content and organizing data. However, GenAI could also be used for information manipulation, posing a more severe threat to democracy.
State intervention and regulation have become topics of widespread discussion in the face of information manipulation’s challenges to democracy. But if not handled carefully, such measures could potentially infringe upon freedom of speech. The need for nuanced and practical considerations in addressing regulation issues has emerged as an essential challenge in managing information manipulation.
This report is the first of a series of three reports. Beginning with the definition of information manipulation, it explores how to achieve a more refined standardization of manipulation techniques to lower the barriers in discussions about information manipulation. The report compares analytical frameworks, such as DISARM, BEND, and RICHDATA, highlighting their advantages and limitations. Ultimately, it selects DISARM, which offers a comprehensive definition of techniques and is regularly updated, as the framework for analysis and subsequent management.
This report uses the 2024 Taiwan presidential election and cases of GenAI being used for information manipulation worldwide as examples to illustrate the potential methods GenAI may employ in manipulating information. Using these examples, it demonstrates how to standardize frameworks for such manipulation techniques, applies them to analyze these methods, and discusses subsequent policy recommendations.
This report finds
- During Taiwan’s 2024 presidential election, cases of information manipulation using GenAI, due to being disseminated from anonymous sources, did not have a meaningful impact.
- The use of GenAI lowers the barrier to information manipulation, increases information production, and consequently imposes a significant burden on fact-checking and response processes.
This report recommends
- In the future, it is essential to implement the responsibilities of platforms and GenAI companies. Platforms should regularly publish relevant data on how they handle information manipulation. Additionally, companies developing Gen AI should conduct due diligence to prevent the misuse of their products.
- The government should strengthen the regulation of GenAI technology, raise public awareness, develop GenAI detection tools, and establish unified regulatory standards through international cooperation.
- The government should collaborate with the private sector to mitigate the impact of information manipulation on society.
DSET will continue to refine our policy recommendations by considering the responses from the Taiwanese government and civil society, as well as subsequent regulatory suggestions. This study aims to standardize and operationalize policy discussions through a framework, hoping to facilitate more nuanced management measures without infringing on freedom of speech, thereby establishing a foundation for democracy and freedom.
