Generative AI makes you more creative
But it makes us all less so.
The Intelligent Friend - The newsletter that explores how AI changes the way we think and behave, only through scientific papers.
Intro 🖍️
HI IF readers, how are you? I am really excited to bring you today's issue, where we will talk about a topic that has been in the daily discussions of the vast application of AI (and in particular Generative AI) in daily life for a while now: AI and creativity. Are there evidences from scientific research that already make us intuit something? The answer is yes, and today we will explore a really fascinating one.
A little curiosity before starting: one of the authors of the paper, Prof.
, is also an author of a newsletter on Substack. You can find his newsletter here, where he shares insights on the impact of AI on work and organizations.
The paper in a nutshell 🔦
Title: Generative AI enhances individual creativity but reduces the collective diversity of novel content. Authors: Doshi and Hauser. Year: 2024. Journal: ScienceAdvances. Link.
Main result: AI-assisted outputs are perceived as more creative, but they tend to be more similar to one another compared to those created solely by humans.
How inspired do you feel?
Think about the latest uses you've made of AI. Maybe you've used software to analyze data. Generated automatic subtitles for your videos. Loved Spotify or Netflix recommendations. Translated a document. The uses are very varied. However, among them, it's difficult not to have brainstorming or idea generation.
Maybe you asked ChatGPT: "What can I get for Christmas to my partner who's obsessed with Wes Anderson?";
Or you wrote on Gemini: "I have some chickpeas, a steak and tomatoes, what recipe can I cook?";
Or you interacted with Claude by writing "Tell me 5 ideas for a new newsletter on famous writers".
All of these tasks have one thing in common: they are creative tasks.
AI, and in particular Generative AI, is increasingly being used for creative tasks and to improve our capabilities in that regard1. And, of course, the tasks do not stop at writing or content ideas, but range from recipes to complex work-related activities, such as "ways to improve this presentation".
This, one can say, has positive effects at first: if I have more ideas to write my novel, if I have more ideas for my song, if I have more hypothetical structures for my presentation, I could intuitively say that my output will be positively affected at least in the initial phase by this broadening of horizons2.
However, there could be downsides too. First, we could anchor ourselves to the ideas suggested by ChatGPT (or Gemini, Claude or other chatbots you prefer). The second is, as someone could suggest, that if we all become more creative, on average it is difficult to predict how much we will have improved overall or not. Or if, in any case, there will be a marked, effective and enjoyable distinction between the various outcomes.
To sum up, using Generative AI tools in a conscious, effective and directed way - and leveraging one's own experience - could have very positive results. But we also need to be aware of the possible outcomes and the various types of fallout.
That is why, in today's paper, the authors focus on how Generative AI can actually make us more or less creative. Let's find out how.
An "augmented" story
As I anticipated, there are several tasks that we could observe to evaluate the impact on "creativity" of the use of chatbots like ChatGPT (used in the study). In previous studies, for example, writing resumes, some employee routine tasks and policy design have been evaluated, among others.
In this research, Doshi and Hauser focus on a task in which more or less all of us have been involved at least once: writing stories. And, in particular, written micro-stories (therefore very short). We have always written short stories, especially if we have a newsletter. Maybe you have used an anecdote to introduce a topic, a story to explain a difficult concept or simply to set the "mood" of the piece. Furthermore, this creative activity has already been used in the scientific literature previously to evaluate differences in individual creativity3.
Having specified the scholars' vision on the task and the objectives, the most attentive readers will not have missed a question. If we want to understand how much people improve their creativity with Generative AI, we need to measure this factor. How do we measure creativity? I think this is a very interesting aspect.
The authors focus on a conceptualization based on two dimensions45:
Novelty: it represents how much a story departs from common expectations or traditional ideas. It was measured through a novelty index, which includes assessing the story’s originality, uniqueness, and rarity;
Usefulness: it's the story’s relevance to the intended audience, its feasibility to be expanded into a complete book, and the likelihood of a publisher being interested in developing it further. Simply put, it was measured by assessing the story’s potential to become a publishable product.
To assess the differences in creativity, the authors followed a rigorous process:
First, they recruited almost 300 participants to write an 8-sentence story for a teenage/young adult audience, with full creative freedom.
Writers were randomly assigned to one of three groups: no AI, AI providing one idea, or AI providing up to five ideas (via GPT-4). Writers in AI groups could request 3-sentence ideas for inspiration.
Finally, they recruited 600 evaluators to rate six randomly selected stories each on creativity, usefulness, and emotional impact, unaware of AI involvement.
A particular improvement
When we think about writing a short story today, it is almost instinctive for some to think about using AI to help us generate ideas. But the yield of a “more creative” story, as we have seen, does not depend on a single factor. Its novelty and its usefulness could go in different directions.
Furthermore, there is another element to consider. An element that someone has thought of so far, but that has not primarily attracted your attention in this discussion. Imagine that you are in a room writing a story using Generative AI, and that in another room there is another person using it. Let's imagine, for example, that you do not have much experience in writing stories.
Now, you put in all your effort with the support of technology. You generate ideas, you rework them, you look for a unique but enjoyable style. You finish the task. A researcher enters, who informs you that in the other room there is Stephen King, who has just finished the task.
As you can imagine from this simple micro-story, there will potentially be enormous differences between a non-expert writer and one of the most important contemporary writers. This is why the experience factor is crucial in the evaluations that are made, and it is also very relevant in the authors' results.
Results are truly fascinating. The research reveals that, first of all, Generative AI improved stories by making them more creative, well-written, and enjoyable. So, both novelty and usefulness increased, on average, in ratings.
For the second result, let’s go back to the example of you and Stephen King. While you might have been put off by the potential comparison with the great writer, who do you think has gained the most from using technology? If your intuition is yours, you are in line with the results of today’s study.
In fact, less creative writers benefit the most from ChatGPT, with creativity improving by 10–11% and the quality and enjoyment of stories increasing by 22–26%.
But the interesting elements do not end there. Looking carefully at the experimental results, the authors observed that Generative AI helps level the playing field by reducing disparities between writers with varying levels of natural creativity.
Therefore, in other words, there is an "equalizer" effect for less experienced writers. However, there is also initial evidence of "flattening" of the evaluations of the various outcomes achieved with the support of Generative AI.
As for every study, the authors provide further details on the results, possible explanations, reflections made on them and limitations of the study. Furthermore, there are many elements of the methodology that are fascinating and that deserve exploration that, of course, we have not covered in this issue. Especially if you are a newsletter author, this study could really be fascinating and I recommend you - as with all the papers I talk about on The Intelligent Friend - to discover more about this study!
Takeaways 📮
Generative AI improves individual creativity. Writers using AI tools created stories rated as more creative, better written, and more enjoyable, with less creative writers benefiting the most.
There is a risk. While AI boosts individual creativity, it leads to more similar stories across writers, reducing the diversity of creative ideas overall.
Have you used AI?! Readers often assign less ownership to AI-assisted stories, emphasizing the need for transparency about AI involvement.
Further research directions 🔭
How can the definition of usefulness be refined and adapted to accurately measure creativity across various domains and mediums?
How do external motivators like financial incentives or intrinsic motivations impact the integration and effectiveness of generative AI in creative processes?
How might generative AI influence creativity differently when applied to other media, such as visual art or music, compared to writing?
The Highlight 🥷🏻
This is the section where I'd like to highlight the amazing work that several authors do here on Substack, through links to their newsletter or specific pieces I've read. Here are some issues you can't miss reading on this platform. In the last issue I made a list of authors whose work I recommend in general. In the next ones, starting from this one, I will highlight specific issues that I found particularly intriguing.
A great Christmas AI roundup, all to nibble on, by
.
If you are a rock enthusiast, you can't miss this issue, by
.
Practical and conscious reflections and suggestions on how to experiment with Generative AI, by
.
Thank you for reading this issue of The Intelligent Friend and/or for subscribing. The relationships between humans and AI are a crucial topic and I am glad to be able to talk about it having you as a reader.
Has a friend of yours sent you this newsletter or are you not subscribed yet? You can subscribe here.
Surprise someone who deserves a gift or who you think would be interested in this newsletter. Share this post with your friend or colleague.
Epstein, Z., Hertzmann, A., Investigators of Human Creativity, Akten, M., Farid, H., Fjeld, J., ... & Smith, A. (2023). Art and the science of generative AI. Science, 380(6650), 1110-1111.
Yanardag, P., Cebrian, M., & Rahwan, I. (2021, June). Shelley: A crowd-sourced collaborative horror writer. In Proceedings of the 13th Conference on Creativity and Cognition (pp. 1-8).
Wolfradt, U., & Pretz, J. E. (2001). Individual differences in creativity: Personality, story writing, and hobbies. European journal of personality, 15(4), 297-310.
Amabile, T. M. (1982). Social psychology of creativity: A consensual assessment technique. Journal of personality and social psychology, 43(5), 997.
Harvey, S., & Berry, J. W. (2023). Toward a meta-theory of creativity forms: How novelty and usefulness shape creativity. Academy of Management Review, 48(3), 504-529.