Teacher Disclosures of AI Practices in their Work: A Gap Between Actual and Visible AI Use

• In Event: Sociology of Education Section Roundtables
  In Refereed Roundtable (60 minute): Table 16. AI for Teaching and Learning

Abstract

When large language models such as ChatGPT became publicly available, teachers were forced to decide whether and how to incorporate these tools into their work, often in the absence of formal district guidance. While emerging research documents variation in teachers’ AI experimentation and avoidance, less attention has been paid to how knowledge about that use is produced. Drawing on scholarship in the sociology of education and organizations on the gap between actual practice and organizationally visible practice, this study examines how teacher AI use becomes visible—or remains obscured—within schools.

Using 30 semi-structured interviews with staff members at a diverse urban elementary school conducted between 2023 and 2025, I identify three “visibility filters” that shape whether AI use enters formal or informal organizational channels: (1) assessments of professional jurisdiction, (2) relational dynamics with administrators and peers, and (3) moral cost–benefit reasoning around workload, equity, and academic integrity. Teachers routinely experimented with AI for planning, differentiation, and bureaucratic tasks, yet selectively disclosed these practices depending on perceived risk, audience, and legitimacy. As a result, vertical visibility to administrators was limited, even as horizontal visibility among trusted peers expanded.

These findings shift analytic attention from measuring AI adoption to examining the social processes that mediate what becomes legible to policy and research. Observed or reported AI use reflects only those practices that pass through organizational, relational, and interpretive filters at a given moment. For policymakers, this suggests that efforts to regulate or support AI based solely on formal artifacts or disclosure mechanisms may systematically misrepresent practice. AI governance, therefore, must be designed with attention to trust, bounded discretion, and the conditions under which teachers are willing to render emerging practices visible.

Author

• Anderson Neal de Andrade, Rutgers University-New Brunswick