Thoughts.

Thoughts on Combating Disinformation

June 5, 2025

Below are a couple of thoughts on combating disinformation I had after speaking at at European Parliament's Liaison Office in the UK to the European Parliament's Special Committee on the European Democracy Shield Mission to the United Kingdom about disinformation and media literacy.

Young people are not merely passive recipients of information vulnerable to algorithmic influence. They also create, share, and amplify information, thus making them an active participant in the information ecosystem. As they are also the most active users of social media platforms and future decision-makers of public affairs, youth engagement is self-evidently indispensable in shaping democratic resilience.

While top-down regulations are vital to ensuring a safer digital space, bottom-up efforts are equally important because people hate to be told what to believe. Instead, the focus should be on empowering young people to figure out how to choose what to believe, which will then inform how they evaluate, filter and disseminate information responsibly. Efforts in cultivating their sense of responsibility, awareness, and knowledge of information manipulation must also start at a young age given the enduring threats of disinformation.

Moreover, in today’s fast-paced media landscape, people have a short attention span and limited time, while people naturally develop a psychological bias towards certain information sources, so disinformation will always remain a threat. While its persistence may seem inevitable, its impact is not. We must help individuals develop an intuitive reflex to recognize common traits of manipulated information. This instinctive awareness forms the realistic foundation of robust digital literacy and critical thinking.

Thoughts on the Use of Computational Methods

April 18, 2025

Below are a couple of thoughts on the application of computational methods in political science I had after attending the MPSA and PSA Annual Conferences in April 2025.

1. The misuse of computational methods remains common, and proper validation is not always conducted. Since these methods heavily depend on training data and parameter configurations, unvalidated or improperly validated models may produce biased results or a false sense of certainty. Thus, results generated by an LLM, a topic model, a clustering algorithm, or other computational approaches should not be accepted immediately without proper validation.

2. While there may be no one-size-fits-all rule, validation is more than simply assessing, for example, whether the clustering results look 'beautiful' or 'ugly'. It is about understanding the statistical implications behind and also the practical usefulness of the results, which points to the need for both robust technical execution and careful logical reasoning with the consideration of the actual context.

3. Computational methods are ultimately tools, not perfect solutions to all data exploration problems. When these tools are applied properly, they can strengthen our exploratory power. Otherwise, they can be counter-productive. If we hope to do more than a descriptive analysis, we must ask ourselves whether our input actually reflects the underlying opinions or intentions we seek to uncover. This is especially the case when examining government and media texts, which do not necessarily indicate internal strategic thinking or do not necessarily converge. Acknowledging and mitigating such limitations is crucial for minimizing bias and improving analytical rigour.