Introduction to Political Science
UAS BFI Vienna
AI was integrated as both a teaching tool and an object of political-science critique. Students developed theory-based prompts to generate and compare election manifestos and speeches across multiple LLMs while using the same prompt in each LLM. They then reflected on bias, propaganda, hegemony, and the politics of information control contained within the generated manifestos. What sets this practice apart is that in addition to students' reflections, the prompt quality was also assessed, allowing the course to simultaneously tackle course content, improve prompt-writing skills, and build critical AI skills by experiencing and reflecting on the bias behind different LLMs.
Learning outcomes
- Describe and explain the main theories, concepts, and terms of political science
- Identify the main aspects of political-science problems
- Apply theoretical models and develop a critical, analytical way of thinking for interpreting and evaluating daily politics and social structures
- Summarize and interpret scientific texts, and identify and evaluate findings and key statements
Assessment
Documented prompts were a major reference for grading. The lecturer deliberately shifted assessment away from outputs and toward whether prompts were anchored in course theory, whether students could critically reflect on generated outputs, and whether they could explain and compare what different models produced. The lecturer assessed the extent to which the prompt refers to course content, whether it includes original definitions or paraphrases, and whether additional elements — such as the behavior of fictional characters — were incorporated into the prompt. Assessment focused on prompts and critical reflection rather than production quality, which also made access differences less consequential for grades.
Evaluation
The course assessments and final exam functioned as evaluation, since the lecturer examined prompts, reflections, and exam grades to see whether AI helped students acquire the learning outcomes. Regular communication with students also served as a measuring tool — for example, feedback after AI activities and students pointing out difficulties in AI tool usage, which resulted in reduced workload for generating audiovisual media in order to shift the focus towards course content and away from individual AI tools.
- KPIs tracked: Yes — Indicators included: weekly satisfaction captured in a logbook; the quality of prompts and reflections in graded assignments (complexity and structure of prompts, inclusion of course content); and verbal feedback captured after grading.
- Formal institutional evaluation: No
Risk management
Risk management worked mainly through pedagogical rather than technical controls. Students had to disclose AI use, submit prompts, and reflect on outputs from different LLMs to prevent overreliance on AI. To earn a good grade, prompts had to be carefully designed and grounded in course theory rather than based on free-form generation. AI was prohibited during the reflection phase, so most reflections took place in class. Where necessary, the lecturer reduced technical demands — for example, by allowing students to create an audio file instead of a full video. Output quality itself was not graded heavily because differences in access, subscriptions, and prior media expertise made direct comparison difficult. A further risk was that students might use AI for prompt generation so extensively that it became hard to distinguish between students' own work and AI contributions, especially since they were encouraged to use AI to improve prompts; this risk was accepted because of the expected benefits, and will be addressed in future through a controlled environment that logs prompt development.
Challenges
- Insufficient prompting and students' occasional overreliance on AI
- Made prompts a core assessment object, requiring theory-based prompting and AI documentation, and used grading feedback to correct weak practice
- Difficulty assessing effort and authorship, since it was hard to differentiate student input from AI contributions within prompts
- A phased approach is planned for the future, where students formulate their own prompts in class first and improve them afterwards using AI
- Unequal access to paid tools, and the lecturer's own limited experience with some AI applications
- De-emphasised output quality where access was unequal, and flexibly adapted tasks
- Too little time for live reflection after the tasks
- Most challenges of this kind can only be effectively managed through structural changes in the curriculum and the institution: more course time, stronger AI-literacy support classes, more prompt-focused assessment, and additional tutorials run by specialists
Scalability
The design is highly transferable to courses dealing with discourse, ideology, communication, bias, public reasoning, or critical media literacy — and more broadly to any theory-based subject aiming to establish a practical connection. It is also suitable for courses with limited time for in-class discussion or activities, such as lectures. Reusable elements are theory-based prompting, same-prompt multi-model comparison, and assessment through reflection rather than output alone. To scale it well, lecturers would need curated tool lists, extra time for live debriefing, support with prompt pedagogy and assessment, clear institutional guidance on documentation/disclosure, and possibly an AI-literacy foundation offer for students. Flexible and individual support also matters, because technical confidence and access to advanced tools vary strongly across students.