The Challenge to Human Creativity: Can AI Surpass Humans in Divergent Thinking?
...or has it already happened?
Creativity has long been considered a uniquely human trait, central to human identity and achievement. However, the rapid development of artificial intelligence (AI) is challenging this notion. Recent AI systems like ChatGPT and Midjourney can produce high-quality artistic and literary works, raising questions about the differences between human and machine creativity. A new study by Koivisto and Grassini published in Scientific Reports tackled this debate by comparing human and AI creativity using the gold standard test of divergent thinking.
A quick intro to creativity research
But first let’s have a quick look at creativity research... Creativity research aims to understand the characteristics, cognitive processes, and environments that allow some people to produce more creative ideas and products than others. A key focus is divergent thinking versus convergent thinking.
Divergent thinking is a key component of creativity that involves generating multiple novel ideas or solutions to open-ended problems. It is contrasted with convergent thinking which focuses on identifying a single best answer. To assess divergent thinking, tests like the Alternate Uses Task (AUT) are used where people come up with creative uses for everyday objects. For example, they might be asked to generate as many possible alternate uses for a brick. Performance is usually measured by the number of responses and their originality — unusual responses indicate higher levels of creativity.
According to associative theories of creativity, individuals with more flexible semantic networks and ability to make connections between remote concepts perform better on divergent thinking tasks. According to these theories, people who can make remote associations will perform better on tasks that require divergent thinking, like coming up with unique uses for a brick.
Controlled-attention theories also emphasise the importance of executive functions like working memory, inhibition, and mental flexibility in regulating the creative process. These theories suggest that creative people are better at using working memory to hold lots of information, inhibiting obvious responses, and flexibly shifting between different approaches. For example, when generating new ideas, creative people can hold many concepts in mind while inhibiting clichéd responses. They can also flexibly switch between different perspectives or strategies. According to these theories, executive functions help regulate and direct the creative process.
Humans VS AI
Koivisto and Grassini compared 256 humans to 3 leading AI chatbots - ChatGPT3, ChatGPT4, and Copy.Ai. Participants were given 30 seconds to generate creative and uncommon uses for objects like bricks, boxes, pencils and candles. The originality of the responses was scored by semantic distance algorithms, measuring originality of responses, and ratings from human judges blind to the source.
On average, the AI chatbots outperformed humans. For example, on the pencil prompt one AI achieved the highest semantic distance score, while two AIs earned top subjective ratings for uses of a box. However, humans came up with the most highly rated responses for brick, pencil and candle prompts. The AIs rarely produced mundane ideas or nonsense, unlike some humans. One human received the lowest possible creativity rating for suggesting using a pencil to "write things down."
The findings indicate current AI surpasses average human creative production, likely thanks to superior memory and lack of attentional lapses. However, highly creative humans can still compete by making surprising connections between concepts. The newest AI, ChatGPT4, earned particularly high subjective scores, perhaps by combining concepts in more nuanced ways.
This research provides important insights on machine vs human creativity. It suggests limitations in average human performance, like difficulties with executive control or lack of motivation, while highlighting that uniquely human qualities like flexible thinking and surprise still underlie the highest levels of creativity. As AI rapidly advances, further studies tracking machine vs human creativity over time are needed.
There are also open questions about the thought processes behind AI idea generation. Do AIs form new connections or just retrieve combinations that exist in their databases? Testing AIs on problems with no prior solutions could further illuminate this. Larger, more diverse human samples are also needed to fully understand variations in human creative thinking.
What does this mean for creatives?
The findings from this study comparing human and AI creativity have important implications for creative professionals like designers, artists, and writers. As AI systems become more capable of generating novel ideas and works, there is concern that they may replace human creatives in some capacities. However, this research suggests that while AI can surpass average human creative thinking, the most highly creative humans are still able to compete at the highest levels.
Top designers possess skills like conceptual combination, mental flexibility, and depth of semantic knowledge that may be difficult for AI to replicate. Expert designers also rely heavily on intuition derived from experience and sensitivity to human context, needs, and emotions - human qualities that exceed current AI capabilities. However, less experienced designers may be more readily replaced by AI for routine design tasks.
For creative fields like visual art, music, and writing that value novelty and subjectivity, the "surprise" factor of human creativity remains essential. AI can produce technically proficient work but often lacks the rule-breaking experimentation, emotional expressiveness and cultural embeddedness that defines groundbreaking art (Elgammal et al., 2017; McCormack et al., 2019). Therefore, while AI may complement human creatives, it cannot wholly substitute the most visionary people pushing creative boundaries.
As AI creative tools become more widely available, human creatives may increasingly collaborate with machines as partners rather than compete against them (Anantrasirichai & Bull, 2021). Preparing for an AI-integrated future requires identifying abilities like imagination, intuition, and inventiveness that make humans indispensable. Understanding the continuing edge human creativity has over AI can guide how creative fields evolve alongside advancing technology…
I’m curious to hear whether AI has been used to generate new music?
Great post and really made me think. I wonder how much attribution bias might happen when people compare something creative such a clothing designed by AI vs Christian Dior. A really creative AI might not get picked out because of the attribution of creativity being with humans over computers.