Dennis Hüninghake

5/5 (2)

Generative AI: From Helper to Thought Partner?

The first time I heard of an generative AI tool was in my exchange semester in Budapest. A fellow student introduced Chat GPT to me and it felt almost like magic. My fellow student asked ChatGPT to summarize a research article I had been struggling with, and within seconds I had a clear overview. It would have taken me at least an hour. I used it for a few weeks and introduced it to some of my other friends who were amazed aswell. Since then, I have experimented with different tools: text-to-text models like ChatGPT for writing support and easy explanations of difficult exercises, and text-to-image models like DALL·E or MidJourney for creating visuals in presentations.

The most remarkable aspects are the speed and inspiration these tools offer. When preparing for group projects for example, I used ChatGPT to brainstorm as an example to give me an outline for a digital strategie for a company. The ideas weren’t perfect, but they helped my group to get started much faster. In addition to that, image generators helped us visualize concepts that would have been difficult to explain with words alone. In this sense, GenAI acts like an assistant that can take on many tasks and implement them super quickly.

At the same time, the limitations are quite obvious. The quality of the results varies often, the information is sometimes outdated or simply incorrect. I have also found that relying too heavily on AI can affect my own critical thinking, as I am sometimes tempted to accept the first answer rather than question it.

In the future, I would like to see improvements in two areas. First, better integration of reliable sources (imagine if ChatGPT could always generate citations in APA style correctly) and more transparency about how answers are generated and where the information comes from.

How about you? Do you use GenAI more as a brainstorming tool, or do you rely on it for polished results? And should universities encourage students to use these tools or restrict them to protect independent thinking?

5/5 (2)

Do you remember how you would have to buy an entire CD just to hear one song, or spend money on a big cable package just to watch one channel? Spotify and Netflix promised to fix that. With streaming, you get to listen to anything or watch whatever you want at any time you want. This is, at least in theory, the Long Tail effect , websites can make money from niche content as much as from blockbusters and “decoupling,” where you’re billed for what you consume and not a thing extra. But does it really work this way?

Evidence shows not quite. Klimashevskaia et al. (2024) show that algorithms are built on a high popularity bias. Instead of offering everyone greater diversity, they prefer to recommend what is popular. A 2025 study goes one further, showing that platforms not only mirror popularity, they reinforce it, propelling hits even higher (Kowald, 2025).

On the user end, Netflix viewers often feel they’ve got too much choice and still watch the same suggestions over and over (Romero Meza & D’Urso, 2024). On Spotify, playlists dominate. Pachali et al. (2025) found that ending up on the right playlist can make or break a song’s visibility. So while the library is infinite, most attention stays fixed on the same few artists or shows.

I think this raises a big question about fairness. Carnovalini, et al. (2025) argue that fixing popularity bias means balancing efficiency with diversity. Personally, I’d love if Spotify or Netflix highlighted more “hidden gems,” not just the Top 10. It would make the experience more exciting for users and more rewarding for creators.

And what about you? Do you tend to go and look up niche material, or do you stick with the advised? Should Netflix and Spotify do more to promote diversity, even if it doesn’t feel as efficient?

Reference

Anastasiia Klimashevskaia, et al. “A Survey on Popularity Bias in Recommender Systems.” User Modeling and User-Adapted Interaction, vol. 34, 1 July 2024, https://doi.org/10.1007/s11257-024-09406-0.

Kowald, D. (2025). Investigating popularity bias amplification in recommender systems employed in the entertainment domain. In Proceedings of the Fourth European Workshop on Algorithmic Fairness (EWAF’25) (pp. 1–7). Proceedings of Machine Learning Research. https://doi.org/10.48550/arXiv.2504.04752

Filippo Carnovalini, et al. “Popularity Bias in Recommender Systems: The Search for Fairness in the Long Tail.” Information, vol. 16, no. 2, 19 Feb. 2025, pp. 151–151, www.mdpi.com/2078-2489/16/2/151, https://doi.org/10.3390/info16020151.

Meza, Laura Romero, and Giulio D’Urso. “User’s Dilemma: A Qualitative Study on the Influence of Netflix Recommender Systems on Choice Overload.” Psychological Studies, vol. 69, no. 3, 24 Sept. 2024, https://doi.org/10.1007/s12646-024-00807-0.

Pachali, Max J, and Hannes Datta. “What Drives Demand for Playlists on Spotify?” Marketing Science, vol. 44, no. 1, 18 Sept. 2024, https://doi.org/10.1287/mksc.2022.0273.

Do We Think Less When AI Thinks for Us?

25

Please rate this