Developing a Curated Chatbot as an Exploratory Communication Tool for Chemistry Learning

Abstract

In a technology-centric world, leveraging digital tools such as chatbots allows educators to engage students in ways that may be more accessible for both parties, particularly in large lecture classrooms. This report details the development of an interactive web-based chatbot to curate content for writing about chemistry in context. Students were assigned a 500-word paper where they discuss general chemistry concepts through the lens of a timely, sustainability-related topic, i.e., water footprint, carbon footprint, or embodied carbon. Discussed herein are the development of the decision tree, the chatbot's components, and results from the initial implementation in a large lecture general chemistry classroom. Over 78% of the 347 enrolled students (271) used the chatbot over 350 times in the 3 weeks leading up to the assigned due date of the paper. Eighty-three percent of the interactions were captured for further analysis, which showed that 22% of students used the chatbot more than once. Forty-six percent of recorded interactions were used to aid students in developing or refining their idea for the assignment. The curated chatbot technology reported here for writing assignments in chemistry can be readily adapted to other aspects of coursework in chemistry.

Students’ Experience of a ChatGPT Enabled Final Exam in a Non-Majors Chemistry Course

“…Course instructors often worked with students to help them refine their topic to be achievable in the relatively short word limit. A curated chatbot was also available to students to answer questions related to the paper.…”

Physical Chemistry Education and Research in an Open-Sourced Future

“…In other words, with enough educational resources generated by experts in a physical chemistry subfield, a large degree of high quality, reliable, and accessible knowledge will be made freely available to allow for AI tutors that can have conversations with a learner (see Figure for an illustration). By 2050, we can also assume that such AI tools will be proficient in “speaking” in most languages (i.e., reading and responding to prompts in different languages), which will further democratize learning. , Already chemical educators have begun leveraging curated chatbots to help students hone their scientific writing skills - a trend of AI integration that will only continue. Similar to how free and open online resources such as Khan Academy and MOOCs are able to close the learning gap and bring students up-to-speed for university coursework, , we expect expert-generated educational content enhanced by AI tools to serve as an approach to lower barriers for researchers at higher levels of scientific research.…”

“…Chemical educators are already experimenting with using these tools to hone students' scientific writing skills. 6 Educators are also grappling with the ability of AI tools to solve relatively difficult college-level calculus and physics problems. 7 Additionally, many aspects of chemical science research from literature searching, coding, data analysis, and robotic operation are already being accelerated using AI tools.…”

ChatGPT Convincingly Explains Organic Chemistry Reaction Mechanisms Slightly Inaccurately with High Levels of Explanation Sophistication

“…Traditionally, chatbots were rule-based and structured as a dialogue tree that can be used to simulate a backand-forth conversation. 7 It is important to note that rule-based systems do not generate new answers; rather, rule-based chatbots match the user's input to a rule pattern and select a response from a predefined set of responses. 8 In contrast, GAIbased chatbots generate text through a form of statistical inference in which words are selected based on the likelihood of their appearing together in the data from which the LLMs were trained.…”

“…Large language models (LLMs) learn from the patterns and structures of human natural language and can generate text for various tasks. Traditionally, chatbots were rule-based and structured as a dialogue tree that can be used to simulate a back-and-forth conversation . It is important to note that rule-based systems do not generate new answers; rather, rule-based chatbots match the user’s input to a rule pattern and select a response from a predefined set of responses .…”