Reconceptualizing outdoor geoscience learning for the generative AI era: a human-centered instructional model

Abstract

Generative artificial intelligence (GenAI) creates a distinctive problem for outdoor geoscience learning: it can broaden questioning and explanation, yet it can also pre-structure what learners notice and displace their interpretation of field evidence. Existing field-learning frameworks explain how conceptual, geographical, psychological, technical, and social novelty shape participation, but they do not fully explain this redistribution of epistemic work and authority. Extending established accounts of technological mediation to field settings, this Hypothesis and Theory article interprets GenAI as an epistemic reorganization of field inquiry rather than as technical novelty alone. It advances the central hypothesis that GenAI is more likely to augment, rather than displace, learners’ epistemic agency when it is introduced after direct observation and an initial learner interpretation, restricted to generating questions and candidate explanations, and subjected to human adjudication against field evidence. On this basis, the article proposes a six-stage human-centered model: preparing for field and GenAI use; attending to and recording field evidence; constructing an initial learner interpretation; engaging GenAI as a dialogic challenger; conducting human evidence-based adjudication; and completing reflective synthesis and epistemic accounting. Field evidence functions as an epistemic constraint, GenAI outputs remain provisional objects of scrutiny, and learners and teachers retain interpretive authority and responsibility. Six testable propositions specify expected effects of sequencing, initial interpretation, challenger-oriented prompting, explicit adjudication, substitution, and teacher orchestration. Operational criteria for distinguishing support, mediation, and substitution, and a thought experiment at a basalt outcrop, show how the model can be applied and tested. The model offers a falsifiable agenda for research in physical, hybrid, and virtual geoscience field learning while defining conditions under which GenAI use should be limited or withheld.