Using Large Language Models to Engage the Public in Your Writing

Many researchers in Condensed Matter Physics and Materials Science find it extremely difficult to translate their research into a form which is accessible to the public. It is important to realize that only 8% of adults in the USA have undergraduate degrees in a STEM area.  More importantly, 54% of adults in the USA read below a 6^th grade level.  Our typical mode of reporting on our research is to use too many buzz words, arcane language, and concepts (such as band structure or crystal symmetry) which are extremely abstract. This makes writing for the public a very challenging and sometimes time-consuming task. One can use large language models such as ChatGPT or Copilot in two ways:  The first is to take an existing public announcement which is written for the educated public in STEM areas and then use the LLM tool to translate it into a form which is accessible to as broad an audience as possible.  We curate the website FunsizePhysics whose target audience is an enthusiastic middle school student, or someone who graduated from high school. We have found that using a prompt such as “Please take the following text and explain to me what it means as if I were a Middle School Student” can be an extremely effective and fast means for translating extremely technical text in a way which at least begins to be accessible to a middle school student. It is possible to take that translation by the LLM and then edit it into a final form.  On the other hand, if you create something for the public, one can use the LLM to analyze what background is required to understand it.   Our preliminary estimate is that such tools can save nearly a factor of four in the time required to translate something written for the NSF or PhysicsWorld in a way which is accessible and interesting to our target audience.