Caregiver involvement in children's homework, especially in mathematics, significantly influences academic achievement. However, not all students have access to caregiver support. Critical barriers identified in prior research include caregiver demands for knowledge, time, and resources. Furthermore, some homework involvement methods can hinder academic progress, such as when caregivers undermine child autonomy, exert undue pressure, or use teaching methods that conflict with school practices. This phenomenon of variable effectiveness becomes more pronounced when technology enters the educational landscape—a domain where rapid advancements can outpace caregiver capabilities and traditional educational support structures. However, technological advancements have also shifted the focus toward technology-based interventions for caregiver involvement. Despite initial advancements in increasing caregiver involvement through the automatic delivery of reminders, there remains a gap in providing fine-grained, personalized support that caters to individual student needs. In this project, I study educational technology design for more effective and equitable caregiver involvement in technology-based middle school mathematics homework. Specifically, I will present design research findings of a novel parent-facing homework support tool providing tailored recommendations to parents for how they might support their children's homework. I will then present work-in-progress findings from two pilot studies of the tool, including interventions for joint caregiver-student practice goal setting. The presentation will close by reflecting on the ethical affordances and challenges of technology support for mathematics homework.

Conrad is a PhD student at the Human-Computer Interaction Institute (HCII) at Carnegie Mellon University’s School of Computer Science, where he is advised by Vincent Aleven and Ken Koedinger. He is broadly interested in studying the effectiveness of educational technologies and pathways through data science methods. In the context of his doctoral research, he explores novel technologies for equitable homework support involving students and their caregivers.

Given the recent advancements and breakthroughs with large language models (LLMs), student models that rely on LLMs have been increasingly developed and used to assess and facilitate learning. However, most of the LLMs are trained on content written by adults, most likely by native speakers of the language. Therefore, the performance of the models developed based on these LLMs may vary for users with different language backgrounds and levels of complexity in their writing. Coincidentally, a few (but limited) research studies show that the use of language can indeed have an impact on the performance of LLMs. For example, low-frequency aspects of language (e.g., semantic anomalies, complex nested hierarchies, and self-embeddings) reveal limitations in the ability of GPT-3 to understand the subtleties of language. Additionally, responses are less accurate when prompts are ungrammatical. As such, in the current study, I examined the robustness of several LLM-based student models that detect student self-regulated learning (SRL) behaviors in math problem-solving. Specifically, I compared how the models vary in performance when predicting the SRL behaviors between students with high and low language complexity, measured by four linguistic measures.

Jiayi (Joyce) Zhang is a PhD student in the Learning Science and Technologies program at the University of Pennsylvania, advised by Dr. Ryan Baker. Joyce’s research largely involves using educational data mining methods to develop student models that measure and understand how learners interact with digital learning systems, assessing learners’ knowledge level, affect, engagement, and use of self-regulated learning strategies in a fair and ethical way

Online learning environments are not accessible to last-mile learners due to infrastructure constraints. Last last mile learners are those who are farthest away, most difficult to reach, andlast to benefit from a program or service. Prior work suggests that Interactive radio instruction (IRI), a form of educational radio, can facilitate learning gains among last-mile learners in low-infrastructure environments. Additionally, a core component of instructional design is the collection and analysis of student behavior and performance data. However, gathering usage statistics, assessment,  and evaluation data for educational radio in traditional IRI has been difficult on a large scale, especially when deployed outside of formal classroom settings. In light of the instructional design and data collection challenges of educational radio, this work investigates the effectiveness of basic mobile phones to facilitate the collection of monitoring and evaluation data for educational radio. We investigate the impact of pairing live-audio with text-based mobile learning on student behavior and performance. We conduct a secondary data analysis of student log data from an interactive radio instruction course facilitated with radio and basic mobile phones in Uganda. We measure associations between student behavior and performance outcomes. Using the ICAP theory of cognitive engagement as an explanatory framework, we found that self-reported measures of engagement are associated with different levels of performance in formative assessments. However, we found limitations in applying the ICAP framework to observed behavioral measures of cognitive engagement for summative assessments. These results suggest that it is possible to make audio learning more interactive. However, we need to design assessments that require higher levels of engagement while staying within the constraints of the audio-only medium. In closing, this presentation will conclude on a reflection of the role of ethics in the design of data analysis for education.

Ren Butler is a Ph.D. student in Human-Computer Interaction at Carnegie Mellon University. Ren researches methods of making computing and engineering education more accessible to learners from under resourced communities. Ren utilizes learning analytics and design-based research techniques to understand and support the social and cognitive processes that enable learners to persist, gain new skills, and develop positive identities in informal STEM learning.

Integrating an equity lens into the fields of Artificial Intelligence in Education (AIED) and Learning Analytics (LA) is crucial for advancing fairness, inclusivity, and social justice in educational practices and outcomes. By applying an equity lens, researchers and practitioners in the field of education can critically examine how data-driven interventions and technologies impact diverse student populations, particularly those from marginalized or disadvantaged groups. This entails identifying and addressing biases in data collection, analysis, and interpretation to ensure that LA initiatives do not perpetuate existing inequalities or disparities in education. Specifically, through this lens, I employ a multimodal approach to understand group dynamics through verbal, nonverbal, and object-based interactions in collaborative learning using adequate AI technologies. I utilize both verbal and non-verbal data to construct activity maps that illustrate students' interaction patterns and address several research questions, including how a predominant focus on verbal data sources might overlook crucial aspects of student learning and introduce biases, as well as how patterns of interactions with learning materials among group members contribute to understanding the associated power dynamics and the distribution of authority within the group over time.

Hakeoung Hannah Lee is a Ph.D. student in the Science, Technology, Engineering, and Math Education Program at the University of Texas at Austin (M.S. in Statistics, B.A. in Education). Her research examines Learning Analytics in the STEM learning context from a situated and embodied perspective. Hannah has published in various journals and proceedings, including the Journal of Learning Analytics, and her work has received funding from multiple organizations. She has been honored with prestigious awards such as the Talent Award of Korea (presented by the Minister of the Korean Ministry of Education) and the Minister of the Korean Ministry of Science and ICT Award. She is a founding member and director of research for the Society of Technology for Education and Learning Analytics (STELA). For more information, please visit her website: https://www.hakeounghannahlee.com/