Conference Session

DASI Technical Session 2: Artificial Intelligence in Higher Education

Collection

2025 ASEE Annual Conference & Exposition

Authors

Lauren Singelmann, Minnesota State University, Mankato; Jack Elliott, Minnesota State University, Mankato; Yuezhou Wang, Minnesota State University, Mankato; Jacob John Swanson, Minnesota State University, Mankato

Tagged Topics

Diversity

Tagged Divisions

Data Science and Artificial Intelligence (DSAI) Constituent Committee

Capabilities to Perform Specific TasksIntroductionGenerative AI (GAI) tools like ChatGPT and Copilot can quickly prepare polished, fiveparagraph essays and clever limericks about any given topic, but can they multiply seventwo-digit numbers? Or answer a question from the Fundamentals of Engineering exam? Or tellyou what the image in a “connect-the-dots” puzzle is? GAI tools are designed to be able toproduce human-like language responses to given prompts, but performance varies depending onthe nature of each task. To further complicate the evaluation of GAI performance, each tool (e.g.ChatGPT, Copilot, Gemini) has its own process for generating responses, and these processescan evolve rapidly – with success varying across tools

Conference Session

DSAI Technical Session 10: Research Infrastructure and Institutional Insights

Collection

2025 ASEE Annual Conference & Exposition

Authors

Julie M. Smith; Jacob Koressel; Sofia De Jesus, Carnegie Mellon University; Joseph W Kmoch; Bryan Twarek

Tagged Divisions

Data Science and Artificial Intelligence (DSAI) Constituent Committee

textual data. For instance, we recently codedapproximately 10,000 state K-12 computer science standards, requiring over 200 hours of workby subject matter experts. If LLMs are capable of completing a task such as this, the savings inhuman resources would be immense.Research Questions: This study explores two research questions: (1) How do LLMs compare tohumans in the performance of an education research task? and (2) What do errors in LLMperformance on this task suggest about current LLM capabilities and limitations?Methodology: We used a random sample of state K-12 computer science standards. We comparedthe output of three LLMs – ChatGPT, Llama, and Claude – to the work of human subject matterexperts in coding the relationship between each state

Conference Session

DASI Technical Session 2: Artificial Intelligence in Higher Education

Collection

2025 ASEE Annual Conference & Exposition

Authors

Xingyu You, Arcadia University; Wang Wang, Arcadia University; Zhairui Shen; Yanxia Jia, Arcadia University

Tagged Divisions

Data Science and Artificial Intelligence (DSAI) Constituent Committee

identify trends in mental health apps since 2009.Privacy policy documents of mental health apps are also collected for analysis. ChatGPT isutilized to extract privacy-related metrics, such as the percentage of apps that reference privacyregulations like the General Data Protection Regulation (GDPR), the Health InsurancePortability and Accountability Act (HIPAA), and the Children’s Online Privacy Protection Act(COPPA). LLMs and RAG are employed to answer critical privacy and security-relatedquestions from the dataset of privacy policy documents. These questions cover multiplecategories, such as the types of user information collected, details on third-party data sharing,and whether users are given options to opt out of data collection.Results:Our

Conference Session

DSAI Technical Session 7: Natural Language Processing and LLM Applications

Collection

2025 ASEE Annual Conference & Exposition

Authors

Suman Saha, Pennsylvania State University; Fatemeh Rahbari, The Pennsylvania State University; Farhan Sadique, Kansas State University; Sri Krishna Chaitanya Velamakanni, Pennsylvania State University; Mahfuza Farooque, Pennsylvania State University; William J. Rothwell, Penn State University

Tagged Divisions

Data Science and Artificial Intelligence (DSAI) Constituent Committee

→ 𝑞" (𝑝 implies 𝑞) might be illustrated using everyday examples like "If it israining, then we take an umbrella." Adopting microlearning as a structured approach couldsignificantly enhance early computer science education, enabling students to solidifyfundamental principles from the start of their studies.Creating microlearning materials can be time-consuming for educators [42]. However,Generative AI like ChatGPT [25] can streamline this process by generating personalizedsummaries, flashcards, and quizzes tailored to specific subjects. Generative AI has become aprevalent tool for content creation across various sectors today. Daniel examined the impact ofgenerative AI on creating learning videos with synthetic virtual instructors, finding

Conference Session

DSAI Technical Session 1: K–12 and Early Exposure to Data Science and AI

Collection

2025 ASEE Annual Conference & Exposition

Authors

Faiza Zafar, Rice University; Carolyn Nichol, Rice University; Matthew Cushing, Rice University

Tagged Divisions

Data Science and Artificial Intelligence (DSAI) Constituent Committee

services to boost productivity and streamline tasks. Google Scholar,for instance, provides a free database that helps students find scholarly articles, research papers,and other academic resources for their projects [15]. Notion serves as an all-in-one productivityplatform, combining note-taking, project management, and collaboration features, making itespecially useful for group work and managing busy schedules [15]. Grammarly, an AI-poweredwriting assistant, helps students refine their writing by checking for grammar, spelling,punctuation, and style while also offering suggestions for improving clarity and organization[14]. ChatGPT stands out as a powerful tool for homework assistance, test preparation,language learning, and other

Conference Session

DASI Technical Session 2: Artificial Intelligence in Higher Education

Collection

2025 ASEE Annual Conference & Exposition

Authors

Ibukun Samuel Osunbunmi, Pennsylvania State University; Taiwo Raphael Feyijimi, University of Georgia; Lexy Chiwete Arinze, Purdue University at West Lafayette (COE); Viyon Dansu, Florida International University; Bolaji Ruth Bamidele, Utah State University; Yashin Brijmohan, Utah State University; Stephanie Cutler, The Pennsylvania State University

Tagged Topics

Diversity

Tagged Divisions

Data Science and Artificial Intelligence (DSAI) Constituent Committee

AI toolsare trained when using the tool. Some questions were posed to encourage critical thinking, such asexamining the data used for AI training, the reliability of AI outputs, and strategies for fine-tuningAI tools. This reflective process aimed to help participants balance human judgment with AIassistance effectively.Furthermore, the participants were introduced to Bloom's taxonomy as a framework for developingAI literacy [13], progressing from foundational knowledge acquisition to the creation of originalwork (See Figure 1). Practical sessions involved the use of resources like ChatGPT, Scholarly,Elicit, and Consensus as AI as a tutor and for aiding literature reviews and syntheses. Similar AIworkshops have been held by the facilitator

Conference Session

DSAI Technical Session 3: Integrating Data Science in Curriculum Design

Collection

2025 ASEE Annual Conference & Exposition

Authors

Md. Yunus Naseri, Virginia Polytechnic Institute and State University; Vinod K. Lohani, Virginia Polytechnic Institute and State University; Manoj K Jha P.E., North Carolina A&T State University; Gautam Biswas, Vanderbilt University; Caitlin Snyder; Steven X. Jiang, North Carolina A&T State University; Caroline Benson Sear, Virginia Polytechnic Institute and State University

Tagged Topics

Diversity

Tagged Divisions

Data Science and Artificial Intelligence (DSAI) Constituent Committee

required by employers. As more data and analytical methods becomeavailable, more aspects of the economy, society, and daily life will become dependent on data-driven decision-making. Recognizing this shift, the National Academies of Sciences (2018)emphasizes that academic institutions must prioritize developing "a basic understanding of datascience in all undergraduates" to prepare them for this new era [1]. This is particularly crucial forSTEM graduates, who must develop varying levels of expertise in working with data – the abilityto understand, interpret, and critically evaluate data, as well as to use data effectively to informdecisions. The recent emergence of large language models (LLMs) such as ChatGPT, which arebecoming increasingly

Conference Session

DSAI Technical Session 1: K–12 and Early Exposure to Data Science and AI

Collection

2025 ASEE Annual Conference & Exposition

Authors

Sri Krishna Chaitanya Velamakanni, Pennsylvania State University; Suman Saha, Pennsylvania State University

Tagged Divisions

Data Science and Artificial Intelligence (DSAI) Constituent Committee

ethical concerns, biases, andover-reliance on AI, which could undermine critical thinking and equitable access to education.E. Microlearning, AI-Driven Feedback, and Student EngagementAI-generated feedback has also emerged as a key enabler of personalized education. Escalantedemonstrated that AI tools provide concise, actionable guidance, aligning with the principles ofbite-sized learning [28]. Similarly, studies such as KOGI's application in programming educationand insights from ChatGPT in first-year engineering courses emphasize the value of modular,on-demand support in enhancing educational outcomes [29]. These works collectively reinforcethe importance of tailored educational resources, such as microlearning videos, in addressing thespecific

Conference Session

DSAI Technical Session 5: Educational Technology and Innovative Tools

Collection

2025 ASEE Annual Conference & Exposition

Authors

Nandan Reddy Muthangi, University of Toledo; Ananya Singh, The University of Toledo

Tagged Divisions

Data Science and Artificial Intelligence (DSAI) Constituent Committee

. Lundberg, "An introduction to explainable AI with Shapley values," SHAP Documentation. [Online]. Available: https://shap.readthedocs.io/en/latest/example_notebooks/overviews/An%20introduction% 20to%20explainable%20AI%20with%20Shapley%20values.html [9] C. Piech et al., "Deep knowledge tracing," in Adv. Neural Inf. Process. Syst., vol. 28, 2015. [10] A. M. Hasanein and A. E. E. Sobaih, "Drivers and consequences of ChatGPT use in higher education: Key stakeholder perspectives," Eur. J. Investig. Health Psychol. Educ., vol. 13, no. 11, pp. 2599–2614, Nov. 2023, doi: 10.3390/ejihpe13110181. [11] Y. Lu, D. Wang, P. Chen, and Z. Zhang, "Design and evaluation of trustworthy knowledge tracing model for intelligent tutoring

Conference Session

DSAI Technical Session 7: Natural Language Processing and LLM Applications

Collection

2025 ASEE Annual Conference & Exposition

Authors

Alexis Frias, University of California Merced; Shrivaikunth Krishnakumar, San Jose State University; Ayush Pandey, University of California Merced

Tagged Divisions

Data Science and Artificial Intelligence (DSAI) Constituent Committee

159 of the 183 Python projects in total,which were divided between training and testing. One student in the dataset developed a longproject that did not fit the maximum prompt size and we had to remove this project’s grading fromthe finetuning. This is an area of improvement for future designs of autograders.2.7 Off the shelf LLMWe conducted experiments using two popular off-the-shelf LLMs: Claude and ChatGPT. We usedsystem messages to enforce a consistent structure across the models, ensuring uniform outputformatting that can be easily parsed into the gradebook Table 7. Table 6: Feedback Comparison between Human and Models system content user content assistant content ### Instruction

Conference Session

DSAI Technical Session 3: Integrating Data Science in Curriculum Design

Collection

2025 ASEE Annual Conference & Exposition

Authors

Ashraf Badir, Florida Gulf Coast University; Ahmed S. Elshall, Florida Gulf Coast University

Tagged Divisions

Data Science and Artificial Intelligence (DSAI) Constituent Committee

, and more time allotted towards the special topics (Machine learning and GeeMap API). I would shorten the panda lesson and python programming lesson by one lesson each to hit another subject in there. It moved well. I would say extend class hours in general for subjects like Matplotlib, Xarray, Numpy, and Pandas so we could cover more topics in the future. I feel that a lesson just on AI coding assistance is not entirely necessary, especially because much of the utility of having an integrated AI API key in the jupyter notebook can be replicate by simplying accessing ChatGPT, Copilot, or any other LLM online. Additionally, it may be worth reducing the lessons on Matplotlib and instead teaching it alongside other lessons. None. I wish we were