From Cognitive Relief to Affective Engagement: An Empirical Comparison of AI Chatbots and Instructional Scaffolding in Physics Education

2025

Read From Cognitive Relief to Affective Engagement: An Empirical Comparison of AI Chatbots and Instructional Scaffolding in Physics Education (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2508-06254.txt.

Introduction

Becker and co-authors compare ways to help ninth-grade physics students in a 2025 field study from Germany. Giving each student tailored help is hard, and tiered hint systems take much staff time to build. The study asks whether a custom AI chatbot can ease mental load and lift mood as well as tiered hints can.

Classes with 273 students in all solved a buoyancy problem with a chatbot, tiered hints, or plain text help. Both live forms of help cut mental load more than text did. The chatbot gave the widest mood gains, with more hope, interest, and faith in one's own skill.

Catalog record

YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE63 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2508-06254

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

From Cognitive Relief to Affective Engagement: An Empirical Comparison of AI Chatbots and Instructional Scaffolding in Physics Education Abstract Providing effective, personalized support is critical for helping students overcome conceptual difficulties in physics. However, established scaffolding methods, such as structured tiered support, are often too resource-intensive for widespread implementation. Therefore, this study, investigates whether an easily adaptable, custom-configured AI chatbot can offer comparable affective benefits and cognitive relief. We conducted a quasi-experimental field study with 273 ninth-grade students in Germany. Classes were randomly assigned to solve a buoyancy problem using one of three conditions: an AI chatbot, a tiered support system, or traditional textbook-style explanations. We measured intrinsic and extraneous cognitive load and affective outcomes (enjoyment, hope, hopelessness, self-efficacy, situational interest) via research-validated questionnaires. Results revealed that both interactive support systems – the custom-configured AI chatbot and tiered hints – were significantly more effective than the textual support in reducing students’ intrinsic and extraneous cognitive load. Furthermore, the AI chatbot yielded the most comprehensive affective benefits, demonstrating significant improvements across all measured affective dimensions, when compared to the textual support. While the chatbot consistently trended more positively than the tiered hints on affective measures, these differences were not statistically significant. These findings suggest that while structured guidance is key to managing cognitive load, the interactive and social nature of AI chatbots holds unique potential for simultaneously fostering positive affective experiences, marking a promising direction for developing effective and holistic learning support tools in physics education. From Cognitive Relief to Affective Engagement \authorsnames [1, \dagger ,1, \dagger ,1, \dagger ,2,3,1]E. Becker, J. Wünsche, J. M. Veith, J. Schrader, P. Bitzenbauer \authorsaffiliations Institute of Physics Education, Leipzig University, Leipzig, Germany, L3S Research Center, Leibniz University Hannover, Hannover, Germany, CAIMed – Lower Saxony Center for Artificial Intelligence and Causal Methods in Medicine, Hannover, Germany † These authors contributed equally \authornote \addORCIDlink J. M. Veith0000-0002-2147-0349 \addORCIDlink E. Becker0009-0009-0985-6335 \addORCIDlink J. Wünsche0009-0007-4380-3590 \addORCIDlink J. Schrader0000-0001-9473-4635 \addORCIDlink P. Bitzenbauer0000-0001-5493-291X Correspondence concerning this article should be addressed to J. M. Veith. E-mail: [email protected] keywords physics education, AI chatbot, cognitive load, motivation, achievement emotions, affective outcomes, instructional scaffolding 1 Introduction Physics is widely recognized as a conceptually challenging subject. A primary difficulty for students is translating theoretical principles into concrete conceptual models, a skill that typically requires deliberate pedagogical support ( Maries et al. 2017a ) . Without such guidance, many novices resort to formula-based problem solving rather than developing a deep understanding of the underlying physics ( Maries et al. 2017a ; Belland 2017a ) . To address this challenge, Differentiated Instruction (DI) is a well-established approach designed to tailor teaching to the diverse needs of all learners ( UNESCO 2020a ; Tomlinson 2014a ) . A pedagogical strategy for achieving DI is employing tiered activities ( Tomlinson 2017a ) , a form of instructional scaffolding, which provides temporary, adaptive support to help a student succeed in a task that would otherwise be out of reach ( Belland 2017a ; Wood et al. 1976a ) . In physics education, “tiered support” (in German: gestufte Lernhilfen) exemplifies this by providing students with a sequence of optional, increasingly explicit learning aids, allowing them to individually adapt the level of assistance they receive ( Stäudel & Wodzinski 2010a ) . The effectiveness of such scaffolding approaches is well-documented, positively impacting not only academic achievement but also student motivation ( Franke-Braun 2008a ; Hänze et al. 2007a ; Schmidt-Weigand et al. 2008a ) . However, widespread application poses a significant practical challenge. While teachers recognize student heterogeneity, they are often constrained by high workloads and limited time ( Robert Bosch Stiftung 2024a ) . This makes the extensive effort required to develop high-quality tiered materials for every task a prohibitive undertaking. Generative AI chatbots, especially when pedagogically tailored through specific configurations, present a promising technological solution to this pedagogical dilemma. They have the potential to act as personalized tutors, capable of reducing teacher workload while offering scalable, on-demand support to students engaged in parallel, individualized learning (cf. Neumann et al. 2024a ; Liang et al. 2023a ) . In fact, current research increasingly demonstrates the potential of AI chatbots to enhance both cognitive and motivational outcomes in STEM education, acting as intelligent tutors or providing personalized scaffolding ( Wang & Fan 2025a ; Chen & Chang 2024a ; Ji et al. 2023a ; Ng et al. 2024a ) . Other studies have explored the application of generative AI in open-ended tasks such as programming, creating teaching aids, or design thinking ( Johnson et al. 2024a ; Huang et al. 2024a ; Ji et al. 2023a ; Yilmaz & Karaoglan 2023a ) . However, it remains an open question how affective outcomes, specifically achievement emotions ( Pekrun 2006a ) , self-efficacy ( Bandura 1978a ) and interest ( Hidi & Renninger 2006a ) are affected alongside cognitive load ( Sweller 1988a ) when using an interactive AI chatbot for physics problem solving, particularly in direct comparison to established pedagogical support systems. Specifically, the distinct effects of dynamic, interactive AI assistance versus more static aids, such as structured tiered hints or traditional textbook-style explanations, on students’ affective and cognitive experiences are not yet well understood. This study directly addresses this gap by systematically comparing the effects of AI chatbot assistance with those of tiered support and traditional textbook-style explanations. To contextualize this comparison, the following section establishes a theoretical framework by examining different forms of instructional support. 2 Research Background 2.1 Support in physics education A central concept in physics education is instructional scaffolding: temporary, interactive support that helps a student build on prior knowledge to complete tasks that would otherwise be out of reach ( Belland 2017a ; Wood et al. 1976a ) . Crucially, effective scaffolding is not merely the provision of information but a dynamic process. Support is offered contingently - in direct response to a student’s progress - and is gradually faded as the learner’s competence grows ( Collins et al. 1989a ; Wood et al. 1976a ) . This responsive nature distinguishes scaffolding from passive instructional aids. A textbook, for example, is a static resource that cannot adapt to a learner’s immediate needs and therefore does not qualify as scaffolding ( Belland 2017a ) . Thus, instructional support can be viewed on a spectrum of interactivity and adaptivity. To evaluate the potential of new, highly interactive systems, it is necessary to compare them against established support methods, across this spectrum, beginning with the most traditional form of assistance. 2.1.1 Textual Support Traditional textual support, most commonly embodied in textbooks, remains an important instructional tool in science classes ( Ogan-Bekiroglu 2007a ; Jonas-Ahrend et al. 2023a ) . In many science courses, such written materials are a fundamental requirement, and instructors expect students to engage with them to enhance academic performance ( French et al. 2015a ) . They are designed to be accessible resources that students can utilize according to their individual learning styles ( Ogan-Bekiroglu 2007a ) . However, a significant discrepancy exists between the presumed importance of textual resources and their actual use by students. Student-reported compliance with assigned readings is varied, ranging from approximately 20% to 70% ( Burchfield & Sappington 2000a ) . This

Read the complete text