Abstract:

Students’ everyday experiences often lead to intuitive explanations of physical phenomena that differ from formal scientific models. When new information appears consistent with learners’ prior ideas, these conceptions may be reinforced rather than challenged, making conceptual change a central goal in science education. In physics education, misconceptions about gravity remain particularly persistent and may hinder the development of scientifically accurate reasoning. Recent developments in Artificial Intelligence (AI) offer new opportunities to support students’ conceptual understanding by providing interactive guidance during independent study. This study investigates the potential of an AI-supported learning assistant to facilitate conceptual change in students’ understanding of gravitational concepts.

Diagnostic assessments play an important role in science education by helping to identify students’ misconceptions and inform the design of effective instructional interventions. Among the available approaches, multi-tier diagnostic tests provide deeper insight into students’ reasoning processes by combining conceptual questions with justification and confidence measures, thereby reducing the influence of random guessing. In the present study, a five-tier diagnostic test was developed based on the Newtonian Gravity Concept Inventory (NGCI). The NGCI is a validated instrument constructed through stages of planning, construction, quantitative analysis, and validation. In this study, the original instrument was adapted into a five-tier multiple-choice format to capture students’ conceptual reasoning, confidence levels, and potential misconceptions. The resulting instrument consists of 25 items designed to assess students’ understanding of fundamental gravitational principles.

The study involves 23 eleventh-grade students (aged 16–17) from a Lyceum in the Famagusta district, Cyprus. Prior to instruction, the five-tier diagnostic test was administered as a pre-test to identify students’ initial conceptions and misconceptions about gravity. During the learning phase, students used Study Buddy (SB), an AI-driven conversational study assistant designed to support independent learning through natural language interaction, to assist with homework and self-study activities related to gravitational concepts. SB provides several pedagogically oriented tools, including concept explanations, text summarization, key-point extraction, self-assessment quizzes, and automated feedback on exercises. Through these features, students can review learning materials, explore scientific concepts, ask questions, and monitor their learning progress through an integrated dashboard environment.

Preliminary analysis of the pre-test responses reveals several recurring misconceptions, including misunderstandings about the relationship between gravitational force, mass, and distance, the conditions under which weightlessness occurs, and the interpretation of gravity as a property emitted by a single object rather than as a mutual interaction between masses.

The study is ongoing, and the diagnostic test will be administered again following the AI-supported self-study period. Comparison of the pre- and post-test results will allow the investigation of potential conceptual change and provide insight into how AI-assisted learning environments can support students in overcoming persistent misconceptions in physics.

Keywords:

Gravity, conceptual change, AI-supported learning, five-tier diagnostic test.