Welcome to the IANUS Case Studies
General event introduction
In this event, the team of the Aristotle University of Thessaloniki (Data and Web Science Lab–Datalab) will explore public attitudes and trust in Artificial Intelligence (AI), focusing on Generative AI, such as ChatGPT.
What follows is a series of visualizations of political discourse during Greece's latest pre-election period. The goal is to analyze the pre-election campaign speeches of Greece’s main political leaders: what issues they raise in the public discourse, what sentiments they convey, and to what extent polarization and populism can be detected in their rhetoric.
To perform the analysis, we used the ChatGPT API and engaged in a programmatic chat with the interactive AI model. Specifically, we prompted ChatGPT to provide us with various linguistic features for each paragraph of the campaign speeches, the results of which we present in the following visualizations.
Before we start we'd like you to fill in your given user ID and answer the following questionnaires.
I1 - Interaction 1
Considering the main objective of this study, along with the information and the model provided by the tool, the key question you need to answer is …
How much do you trust ChatGPT to analyze political discourse?
How to read this plot
What percentage of each speech is devoted to presenting each party's programmatic agenda, and what is the extent of the opponents' criticism?
You may click on the word “Agenda” to see the topics each party leader chooses to discuss. By selecting the word “Criticism”, you’ll navigate the topics they criticize their opponents about.
Do you find the above analysis sufficient or would like to read the speech yourself?
I1 - Interaction 1: Questions
I2 - Interaction 2 - Questions
In this task, two plots (visualizations) representing two different approaches of analysis are provided, and you are asked to evaluate both.
How to read this plot
The step charts below depict the level of polarization (i.e., the intense division of society into opposing ideological camps, fostering exclusion of alternative views and emphasizing emotional and identity-driven narratives) during a political speech, with the number of paragraphs shown on the horizontal axis. Each paragraph, depending on the value of the polarization indicators it has received, is placed in a category:
- no or low level if it has a value from 0 to 0.5,
- medium level if it has a value from 0.51 to 0.8,
- high level if it has a value from 0.81 to 1.
If you wish, hover over the chart to read the corresponding paragraph.
I3 - Interaction 3 - Questions
The results visualized in this plot are initially produced by the ChatGPT, but they are further analyzed and validated by data journalists from the Incubator for Media Education and Development (iMedD).
iMedD is a non-profit organization with a mission to support transparency and independence in journalism and promote meritocracy and excellence in the field.
How to read this plot
The step chart below depicts the level of polarization during a political speech, similar to the previous task.
Submit your answers
Thank you for your participation in the IANUS case studies.
Please submit your answers!
