Navigating the Fine Line: Balancing AI-Generated Creativity and Authenticity in Science Communication
The rise in popularity of general purpose artificial intelligence programs have sparked concerns on the potential misuse in many fields, including writing. Incorporating AI into one’s creative process is like walking on thin ice. The concern of loss of creativity and the inability to validate authenticity is a broadly debated topic.
Creativity: humans vs AI
There is no consensus on the risk and benefit that generative AI would bring. Some researchers found that AI is more creative at certain tasks, such as the Divergent Creativity Task where a participant or program is asked to name ten nouns that are unrelated; the creativity of such task is deemed to be higher the further away the nouns differ from each other.12 Others believe that the most creative kinds are still among us, that despite AI having a more consistent performance measured by quantified creativity score, human participants always had the highest scores in all trials.3 Despite the results, it is still too soon to determine whether AI has the ability to catch up to human creativity. Creativity is deeply engraved into many aspects of our day to day life, and one quantitative measurement of creativity doesn’t capture the full extent of our creative potentials.
Let’s talk honesty
In a time when detection tools for generative AI have not matured yet, the embodiment of authenticity in creative works relies almost solely on the honest disclosure from the works’ producers. Many forms of self or automated disclosure methods have emerged in the past year or so: camera manufacturers have started implementing watermarks embedded in the files to distinguish real pictures from generated counterparts4; Instagram started allowing users to tag their posts as “made by AI” (although the platform itself seems to have randomly tagged non-AI works as AI? we’ll discuss more on this in a moment)5; the Not By AI badge started popping up on websites that choose to distant themselves from publishing generated pieces.
Instagram’s tagging approach is merely a yes or no choice on the involvement of AI. In reality, one may choose to employ AI in just a part of their workflow, blurring the line of whether a piece is fully AI-generated or not.
Adobe’s Photoshop has introduced AI-based background removal and substitution tools, replacing the often tedious process of selecting a part of the image using lasso tools. There is also Generative Fill, developed for the latest releases of Photoshop by Adobe, that adds illusive content to a picture; an artist could totally take a picture with their real camera, put the picture into Photoshop, and start telling the program to add some non-existent imagery to it. Does it count as AI generated artwork? Not necessarily. But it’s also not completely real.
A writer may choose to use the latest version of AI-infused Grammarly to improve their grammar and writing style. The organic content written by a real person shouldn’t be undervalued, but does the use of Grammarly AI suddenly make the work unworthy of an authenticity badge?
The mix of AI-assisted functions into a creative workflow leaves room for interpretation on what really means for something to be generated by AI. Currently, Not By AI requires a 90% estimation of work being done by a person and not by a program.6 Before a more rigid definition on what is generated and what is not can be agreed upon by most people, the authenticity of any kind of creative output relies heavily on the honest self-disclosure by the authors and artists alike.