AI Sensing and Intervention in Higher Education: Student Perceptions of Learning Impacts, Affective Responses, and Ethical Priorities

Paper Title

AI Sensing and Intervention in Higher Education: Student Perceptions of Learning Impacts, Affective Responses, and Ethical Priorities

Publication Info

• Topic area: AI sensing and intervention systems in education
• Keywords: AI in education, student perceptions, ethical concerns, autonomy, privacy, gaze-based detection, facial emotion detection, learning impacts, affective responses

Background and Problem

• Problem / challenge: Current AI sensing-intervention systems in education focus on technical performance but often neglect student-centered factors such as affective experiences, autonomy, and ethical concerns.
• Significance: Understanding student perceptions is crucial for aligning AI systems with their needs and values, ensuring effective and acceptable educational technologies.
• Motivation and related work: Prior studies have highlighted technical benefits of AI sensing-intervention systems but lack empirical evidence on student affective and ethical responses, leaving gaps in understanding how these systems impact learning experiences and well-being.

Solution

• Proposed approach: Mixed-method experimental study with Australian university students to evaluate perceptions of AI sensing-intervention systems and prioritize ethical concerns.
• Novelty:
  1. Empirical evidence on student affective and pedagogical responses to AI sensing-intervention systems.
  2. Ethical prioritization framework based on student perspectives.
  3. Design implications for student-centered, ethical AI systems in education.
• Procedure and key techniques:
  • Stage 1: Scenario-based experiment with six video prototypes varying sensing modality (gaze-based vs. facial emotion detection) and intervention form (system-generated vs. teacher-mediated).
  • Stage 2: Ethical prioritization tasks using Likert-scale ratings and pairwise comparisons of six ethical principles (privacy, autonomy, fairness, accuracy, transparency, learning beneficence).
  • Analysis: Quantitative (Mann–Whitney U tests, Bradley-Terry Model) and qualitative (content analysis) methods.

Results

• Concrete findings:
  • Students rated AI sensing negatively, with significant discomfort and anxiety regardless of sensing modality.
  • System-generated hints were preferred over teacher-mediated interventions due to concerns about agency and social embarrassment.
  • Privacy and autonomy were ranked as the most important ethical concerns, while fairness and transparency were less prioritized.
• Advantage over baselines:
  • Highlights student-centered ethical priorities and affective responses, challenging technology-centric evaluations of AI systems.
  • Offers actionable insights for designing less intrusive, more autonomous AI interventions.
• Experiments / evaluation:
  • Participants: 132 Australian university students.
  • Metrics: Beliefs about learning, affective responses, ethical prioritization.
  • Tools: Video scenarios, Likert-scale ratings, pairwise comparisons.
• Limitations and future work:
  • Limited to hypothetical scenarios; future studies could use real systems for richer insights.
  • Focused on Australian higher education; broader studies across cultures and age groups are needed.
  • Long-term effects of AI sensing-intervention systems remain unexplored.

Summary

This study examined student perceptions of AI sensing-intervention systems in higher education, revealing negative affective responses and ethical concerns about privacy and autonomy. Students preferred system-generated hints over teacher-mediated interventions, emphasizing the importance of agency and avoiding social embarrassment. Ethical priorities focused on autonomy and privacy, outweighing perceived learning benefits. The findings provide actionable design implications for creating non-intrusive, student-centered AI systems that respect ethical values and enhance learning experiences. Future research should explore real-world implementations and long-term impacts across diverse contexts.