Course philosophy

Today, there are plenty of ways to learn about GenAI. What, I believe, sets this course apart is that it‘s based on discussing research papers. To make this statement less abstract, let me illustrate the course philosophy with a specific example.

In many learning materials, you might encounter advice to assign a specific persona to GenAI. For example, if you want ChatGPT to help you write an email to resolve a difficult situation, you might start your request by saying “You are a well-known negotiator.” If you’re going to ask GenAI to help with text annotation, you might say “you‘re a linguist with 15 years of experience in text analysis.”

The problem with this advice is that it isn’t grounded in evidence. In fact, there is a research paper that systematically examines this strategy and finds that “adding personas does not improve model performance across a range of questionsWhen ”A Helpful Assistant” Is Not Really Helpful: Personas in System Prompts Do Not Improve Performances of Large Language Models (Zheng et al., 2024) . Another paper even suggests that assigning a persona to ChatGPT increases the toxicity of its outputsToxicity in chatgpt: Analyzing persona-assigned language models (Deshpande et al., 2023) .

However, simply dismissing personas by citing these references would be misleading, as the studies don‘t consider all possible contexts. By discussing the papers in detail, I hope to help learners move beyond simplistic statements like “personas are useful” or “personas don’t work” and enable them to form a more nuanced, evidence-based understanding of GenAI.

In the same spirit, this course intentionally avoids providing detailed tutorials for specific AI tools. I believe that once you‘re equipped with a basic understanding of the technology and its opportunities, it’s better to experiment and try different tools on your own to find what suits your particular working style and needs. And given the dynamic nature of the field, any specific tutorial would quickly become outdated anyway.

I aim to make this course as accessible as possible and it shouldn’t require any specialised knowledge to follow. However, I assume that the reader is curious and reflective—qualities which I believe are essential in research.