.png){kind=icon}
.png){kind=icon}
Return
University students’ evolving conceptual metaphors of AI in English academic writing: An actor-network perspective
.png){kind=icon}
PRE
AI
J
D
DOI：10.1016/j.system.2026.104115.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Students' metaphors of AI describe and actively configure learner-AI relationships in academic writing. Prior research has identified recurring framings of AI as tool-based or human-based, yet has largely relied on cross-sectional designs that tended to treat metaphors as static representations. This longitudinal study draws on Actor-Network Theory (ANT) as an interpretive lens to examine how ten university students' metaphorical conceptualizations of AI evolved over a one-year period. Data were collected through two rounds of semi-structured interviews, triangulated with students' AI-use artifacts. The findings reveal three trajectories: maintenance of tool-based metaphors with enhanced recognition of AI capabilities; shifts from tool-based to human-based conceptualizations; and processes of intensified or reduced personification. These shifts of metaphors appear to reflect not individual cognitive change alone but negotiated relationships among tools, teachers, peers, policies, and tasks. Effective and ethical human-AI collaborative writing is less an individual achievement than a network accomplishment, dependent on alignment among student literacies, AI capabilities, institutional policies, and task demands. This study demonstrates the value of an ANT perspective in AI research, showing how human and non-human actants mutually shape learning relationships. Creating opportunities to make students' AI metaphors visible can facilitate responsive institutional policy-making and inform pedagogical strategies. The tensions between students' explicit tool-based metaphors and their implicit affective characterizations highlight AI's emotional dimension, inviting further investigation. The study underscores the importance of longitudinal designs for capturing the dynamic evolution of human-AI interaction.
Journal
IF:
5.6
Papers:
479
Citations:
9.7K
Organization
Cited Papers