Source-linked AI summary
Understanding the Practices, Perceptions, and (Dis)Trust of Generative AI among Instructors: A Mixed-methods Study in the U.S. Higher Education
Wenhan Lyu, Shuang Zhang, Tingting, Chung, Yifan Sun, Yixuan Zhang
TL;DR
Higher education lacks sufficient evidence about how instructors use GenAI and how trust and distrust shape that engagement. Using mixed methods with 178 instructors at one U.S. university, the study finds limited direct instructional use despite familiarity, and shows that trust and distrust can coexist. It concludes that calibrated trust requires informed verification, attention to distrust, and further evidence across settings.
Problem
Research has insufficiently examined instructors’ GenAI practices and the distinct role of distrust alongside trust in higher education.
Method
The study surveyed 178 instructors at a single U.S. university and analyzed quantitative and open-ended responses about GenAI practices, perceptions, trust, and distrust.
Results
Instructors reported familiarity with GenAI but limited direct instructional use, while trust and distrust were related yet distinct and could coexist.
Takeaways & Limitations
Calibrated engagement should combine verification with strategies that address distrust and support instructors’ balanced, ethical integration of GenAI.
Takeaways & Limitations
The single-university sample limits generalizability across countries and educational contexts, and the cross-sectional design cannot show how attitudes and use evolve over time.
Abstract
from arXiv · showhide
Generative AI (GenAI) has brought opportunities and challenges for higher education as it integrates into teaching and learning environments. As instructors navigate this new landscape, understanding their engagement with and attitudes toward GenAI is crucial. We surveyed 178 instructors from a single U.S. university to examine their current practices, perceptions, trust, and distrust of GenAI in higher education in March 2024. While most surveyed instructors reported moderate to high familiarity with GenAI-related concepts, their actual use of GenAI tools for direct instructional tasks remained limited. Our quantitative results show that trust and distrust in GenAI are related yet distinct; high trust does not necessarily imply low distrust, and vice versa. We also found significant differences in surveyed instructors' familiarity with GenAI across different trust and distrust groups. Our qualitative results show nuanced manifestations of trust and distrust among surveyed instructors and various approaches to support calibrated trust in GenAI. We discuss practical implications focused on (dis)trust calibration among instructors.
1 Introduction
GenAI is reshaping higher education while creating uncertainty about how it should be used, regulated, and assessed. This study examines instructors’ practices, perceptions, and the distinct but related forms of trust and distrust shaping calibrated engagement.
- GenAI is transforming higher-education teaching and learning while creating uncertainty for faculty, students, institutions, and policymakers.Stakeholders must balance potential innovation with ethical considerations and academic integrity.
- Existing educational GenAI research has examined students’ and instructors’ perceptions, practices, use cases, risks, and challenges.
- Fewer studies have unpacked instructor trust and distrust as distinct responses to GenAI’s reliability, harms, and alignment with instructional goals.Trust involves willingness to rely on GenAI despite vulnerabilities, whereas distrust can limit reliance or usage.
- Balanced perspectives aim to avoid blind trust and blind distrust, which can contribute to a trust crisis when GenAI is accepted uncritically or rejected without sufficient reasons.
- The study uses mixed methods with 178 instructors to examine GenAI practices, perceptions, trust, and distrust in a culturally specific university context.
- The authors analyze coexisting trust and distrust and propose practical design implications for instructors’ informed, balanced engagement with GenAI.
2 Related Work
Prior work has studied educational technology and GenAI acceptance, especially among students, but instructor-focused research remains comparatively limited. This literature has emphasized benefits and risks more often than the distinct dynamics of trust and distrust across higher-education contexts.
- Technology-acceptance research in education has traditionally focused more on students than instructors, despite instructors’ influence on university technology adoption.
- GenAI offers higher-education applications such as generating course materials and assisting students’ programming tasks while also introducing teaching and learning challenges.
- Research has examined student perceptions of ChatGPT across Asia, Europe, Australia, and North America, identifying shared ethical concerns across cultural contexts.
- Instructor-focused GenAI studies span several regions but mostly emphasize perceived benefits and risks rather than systematically differentiating trust from distrust.
- The literature leaves a gap in understanding how instructors across a broad range of fields relate to GenAI and how trust and distrust manifest among them.
- Trust is associated with willingness to rely on systems, whereas distrust reflects suspicion and expected harm, making both relevant to educational technology engagement.
- The study addresses two questions about instructors’ current GenAI practices and how trust and distrust manifest in U.S. higher education.
3 Method
The study surveyed instructors at one Mid-Atlantic U.S. research university using structured and open-ended questions about GenAI practices, familiarity, attitudes, trust, distrust, and teaching contexts. Quantitative and qualitative analyses were used to examine these responses.
- Survey Study Recruitment & Participants Overview: The researchers surveyed 178 instructors at a Mid-Atlantic U.S. research-focused four-year university in March 2024 after institutional review-board approval.The university offered undergraduate and graduate programs across arts, sciences, business, and law.
- Survey Design: The survey comprised blocks on GenAI practices and familiarity, attitudes and trust, demographics, and open-ended perspectives.
- GenAI Practices and Familiarity: Questions examined current and intended GenAI use across instructional activities, typical task types, and attitudes toward training and readiness.
- Attitudes, Trust, and Distrust: Trust measures addressed efficiency, personalization, adaptive teaching, course incorporation, curriculum change, and student problem-solving.
- Attitudes, Trust, and Distrust: Distrust measures addressed threats to student mental health, misinformation, skill degradation, content accuracy, creativity, and independent problem solving.
- Data Analysis: The analysis examined whether familiarity, trust, and distrust varied across instructors’ teaching levels and other groups.
- Data Analysis: Welch’s ANOVA and ANOVA were used for quantitative analyses, while open-ended responses were analyzed using a General Inductive Approach.The qualitative dataset contained 294 valid responses totaling 10,269 words.
4 Survey Results
Surveyed instructors reported moderate-to-high familiarity with GenAI, but limited direct instructional use alongside substantial interest in training. Their trust and distrust were related yet distinct, coexisting across groups and varying with familiarity and teaching level.
- Familiarity and current practices: 82.6% of instructors reported being very or somewhat familiar with GenAI concepts.Very familiar responses accounted for 29.8%, and somewhat familiar responses for 52.8%.
- Familiarity and current practices: 58.8% had included a GenAI statement in their syllabus, while classroom use emphasized ethical implications, knowledge creation, and general principles.The reported proportions were 40.3% for ethical implications, 38.4% for strengths and weaknesses in creating knowledge, and 39.4% for general principles.
- Familiarity and current practices: 64.0%–75.8% had never used GenAI for surveyed educational tasks, and only up to 2.8% used it routinely for text and code generation.Frequent use ranged from 6.7% to 17.4%, while constant use ranged from 3.9% to 9.0%.
- Training needs: 75.7% said the university needed more GenAI training, while 61.6% said they themselves needed it and 62.2% said students needed it.Additionally, 46.7% lacked a clear understanding of classroom handling, and 55.1% considered their training insufficient.
- Trust and distrust: Trust and distrust formed reliable but distinct constructs, with trust items positively correlated internally, distrust items positively correlated internally, and trust negatively correlated with distrust.The scales showed strong internal reliability, with Cronbach’s α of 0.924 for trust and 0.860 for distrust.
- Trust and distrust: Trust and distrust could coexist: instructors were categorized into four groups, including high-trust-high-distrust, and extreme distrust levels were absent.Familiarity differed significantly across groups (p < 0.05), with higher-trust groups tending toward higher familiarity and low-trust-high-distrust groups showing moderate familiarity.
- Teaching-level differences: Teaching level significantly influenced trust and distrust (p < 0.05): instructors teaching both undergraduate and graduate courses had mean trust of 3.39, versus 2.62 among undergraduate-only instructors.Undergraduate-only instructors showed the highest distrust, while graduate-level instructors showed slightly lower distrust.
- Overall pattern: Overall, instructors showed a gap between conceptual familiarity and hands-on instructional use, alongside demand for clearer guidance and training.Many incorporated GenAI discussions into courses, but few used GenAI in direct instructional tasks.
5 Qualitative Findings
Instructors described distrust through concerns about social justice, environmental and labor costs, academic integrity, and threats to learning. They also identified burdens on faculty and students, while some advocated cautious, guided engagement with GenAI.
- Concerns surrounding Social Justice Issues: Instructors associated GenAI with linguistic, social, and copyright concerns, including reinforcement of dominant norms and existing inequalities.These concerns extended from generated outputs to the broader ways GenAI is deployed in education.
- Environmental and Ethical Concerns: GenAI’s resource demands raised concerns about energy and water consumption, environmental sustainability, and labor conditions in model training.Instructors called for greater understanding of models’ capabilities, limitations, and hidden costs.
- Ethical and Pedagogical Concerns: Concerns about fabricated sources, inappropriate analyses, and students presenting AI-generated work as their own led instructors to reconsider assignments and assessments.Proposed responses included greater reliance on in-class exams and questioning the purpose of assignments that AI can complete.
- Ethical and Pedagogical Concerns: Instructors feared that GenAI could weaken comprehension, writing, critical inquiry, analytical skills, creativity, and independent learning.Some worried that ready-made or overly advanced answers could encourage shortcuts instead of understanding and evidence-based reasoning.
6 Discussion
The discussion interprets instructors’ limited everyday GenAI use alongside substantial interest, complex trust dynamics, and ethical concerns. It proposes evidence-based, context-sensitive supports for calibrated trust while acknowledging unresolved environmental, institutional, and attitudinal challenges.
- Instructor practices and pedagogical implications: Instructors showed positive intent and familiarity with GenAI, but everyday instructional use remained low and many prohibited GenAI in exams.The discussion links exam restrictions to concerns about academic integrity or uncertainty about governing AI-assisted assessment.
- Comparative context: Comparisons with other surveys are constrained because most were conducted in 2023 while GenAI continues to evolve rapidly.The authors recommend reporting data-collection dates prominently to improve contextual interpretation.
- Instructor practices and pedagogical implications: GenAI may support critical thinking, creativity, and engagement through assignments that compare AI-generated solutions with students’ work to identify gaps or biases.This pedagogical use extends beyond feedback and illustrative examples.
- Trust and distrust dynamics: Trust and distrust were distinct, coexisting responses shaped by familiarity, pedagogical alignment, ethical considerations, and concerns about integrity, bias, and the digital divide.Highly trusting instructors often adopted a cautious “trust, but verify” approach.
- Supporting calibrated trust: Calibrated trust may be supported through interactive experimentation, user-friendly interfaces, guided modules, tailored training, sandboxes, and curricula combining technical, ethical, and pedagogical learning.The proposed supports aim to help instructors incorporate GenAI while maintaining academic integrity and pedagogical quality.
- Evidence-based adjustments: The discussion calls for empirical studies testing how training, support, and task complexity affect faculty trust and identifying evidence-based interventions.Controlled trials and pre/post trust measures are proposed for examining trust development among instructors.
- Nuances of trust and distrust: Existing trust-calibration strategies may not address instructors with entrenched distrust or excessive trust, whose views may be resistant to materials challenging their preconceptions.The authors suggest further research on the contextual and ideological roots of these extremes.
- Environmental and ethical challenges: Environmental impacts, including high energy and water use, create ethical tensions that may require sustainability-linked policies and lower-resource AI approaches.The discussion mentions model distillation and edge computing as possible lower-resource solutions.
7 Limitations
The study’s generalizability is limited by its sample and cross-sectional design. Broader and longitudinal research is needed to examine diverse educational settings and changing GenAI attitudes and use.
- Scope and generalizability: The sample size limits generalizability across countries, educational contexts, and stakeholders.Future research should include settings such as K–12 education and other sociocultural contexts.
- Study design: The cross-sectional survey captures one moment and cannot establish how instructors’ GenAI attitudes and use evolve over time.Longitudinal studies are needed to investigate changing dynamics and long-term implications.
8 Conclusion
The study finds that instructors’ familiarity can coexist with limited direct GenAI use, while trust and distrust remain related but distinct. It concludes that future interventions should cultivate calibrated trust and address distrust’s roots.
- Conclusion: Among 178 instructors at one U.S. university, familiarity with GenAI coexisted with limited direct experience and application in educational tasks.The findings describe current practices within a single institutional context.
- Conclusion: Trust and distrust could coexist as distinct concepts, revealing nuanced dynamics that future research should examine together.The study identifies multiple factors contributing to their formation, including cases of blind distrust.
- Conclusion: Many instructors used “trust, but verify” approaches, supporting calibrated trust and distrust in their own and students’ GenAI use.The conclusion calls for interventions that also address the roots of distrust.
Acronyms
The paper defines abbreviations used for GenAI, STEM, ITS, IRB, and TAM.
- Acronyms: GenAI means Generative Artificial Intelligence; STEM means science, technology, engineering, and mathematics.These abbreviations identify the paper’s central technology and an academic disciplinary grouping.
- Acronyms: ITS means Intelligent Tutoring System; IRB means Institutional Review Board; TAM means Technology Acceptance Model.These abbreviations refer to an educational technology, research oversight, and an acceptance framework.
A.1 Theme 1: GenAI Usage
Theme 1 examines surveyed instructors’ current GenAI teaching practices, experiences, and perspectives.
- The theme describes surveyed instructors’ current practices involving GenAI in teaching.
- It captures instructors’ experiences with GenAI in their teaching.
- It also examines instructors’ perspectives on using GenAI for teaching.
A.2 Theme 2: Impact of GenAI
Theme 2 examines surveyed instructors’ concerns about GenAI’s effects on students and instructors.
- The theme focuses on surveyed instructors’ concerns about GenAI’s effects on students.
- It also addresses concerns about how GenAI affects instructors.
- These concerns are organized as the study’s theme on GenAI’s impact.
A.3 Theme 3: Perspectives on GenAI
Theme 3 examines surveyed instructors’ perspectives on GenAI models and their ethical, legal, and practical implications.
- The theme captures surveyed instructors’ perspectives on GenAI models.
- It considers the ethical implications of GenAI models.
- It addresses the legal implications of GenAI models.
- It also covers the practical implications of GenAI models.