Emergence and development of observation practices in AI-assisted plant adaptation learning

Abstract

Observation is a fundamental scientific practice in science learning, particularly for understanding gradual biological phenomena such as plant adaptation. Despite the increasing integration of artificial intelligence (AI) in education, limited research has examined how AI-supported tools influence students' observation practices during authentic scientific inquiry. This study investigates the emergence and development of students' observation practices during plant adaptation learning supported by an AI-based identifier. A six-week mixed-methods exploratory study was conducted with 84 middle school students. Learning activities were organised into four phases: orientation, initial observation, AI-assisted identification, and re-observation and validation. Data were collected through observation sheets, students' observation notes, reflective journals, group discussions, and records of AI use. Quantitative analysis revealed significant differences in observation orientations across learning phases (Friedman test, p < .001), indicating changes in observation practices over time. Qualitative findings showed that students initially relied on AI-generated cues to guide their observations. Over time, however, observation practices shifted toward direct empirical observation and peer-supported validation through repeated comparisons of plant characteristics across environmental conditions. The findings reveal a developmental trajectory from AI-reliant observation to direct observation and collaborative validation, accompanied by a gradual reduction in dependence on AI-generated outputs. These findings suggest that AI may function as an initial observational scaffold, while more sophisticated observation practices emerge through repeated engagement with phenomena and collaborative reasoning.