Reflective AI: A Slow Technology Approach for Design Education

Paper Title

Reflective AI: A Slow Technology Approach for Design Education

Publication Info

• Topic area: Integration of slow technology principles into AI-based design education.
• Keywords: Reflective AI, slow technology, design education, object detection, critical reflection, creative practice, AI literacy, annotation, design pedagogy, machine learning.

Background and Problem

• Problem / challenge: Current AI tools in design prioritize efficiency and rapid output, undermining critical reflection, creativity, and agency. These tools obscure the underlying processes, limiting students' ability to critically engage with AI technologies.
• Significance: Addressing this issue is crucial for cultivating reflective practices and critical technical literacy in design education, enabling students to navigate the increasing role of AI in creative fields.
• Motivation and related work: Prior work has explored AI's socio-technical dimensions and reflective practices but has not adequately addressed the challenges posed by "fast AI" tools in design education. Existing approaches lack structured methodologies for integrating AI as a reflective medium rather than a productivity tool.

Solution

• Proposed approach: Reflective AI, a slow technology framework that treats AI systems as mediums for critical reflection rather than tools for rapid output.
• Novelty:
  1. Development of a structured methodology for implementing Reflective AI in design education.
  2. Empirical evidence showing how slow engagement with AI components fosters reflection and technical understanding.
  3. Introduction of material and temporal disentanglement as core mechanisms for Reflective AI practice.
• Procedure and key techniques:
  • A three-day workshop called the Objective Portrait Workshop was designed for design students.
  • Students engaged in dataset creation, annotation, model training, and interpretation of AI outputs.
  • Tools like the Quadrant Tool were used to structure label selection and foster subjective reflection.
  • The process emphasized slow, deliberate engagement with AI components to cultivate critical and reflective practices.

Results

• Concrete findings:
  • Students developed deeper self-awareness and critical understanding of their creative processes.
  • They gained practical insights into AI principles, such as generalization and annotation consistency.
  • Reflective engagement led to symbolic and metaphorical interpretations of AI outputs, enhancing creative exploration.
• Advantage over baselines:
  • Unlike "fast AI" tools, Reflective AI fosters critical reflection, agency, and a deeper understanding of AI systems.
  • The workshop enabled students to see AI as a reflective medium rather than a black-box tool.
• Experiments / evaluation:
  • Conducted with 11 design students over three days.
  • Activities included dataset preparation, label selection, Object Portrait creation, and AI interpretation.
  • Data collected included annotated datasets, Object Portraits, reflection cards, and tutor notes.
  • Thematic analysis was used to identify patterns in student reflections.
• Limitations and future work:
  • The study was limited to design students and object detection models (YOLOv5).
  • Future work could explore Reflective AI with generative models and extend the methodology to professional design contexts.
  • Additional research is needed to integrate Reflective AI into broader design education curricula.

Summary

This paper introduces Reflective AI, a slow technology approach that reframes AI systems as reflective media for design education. Through a structured three-day workshop, students engaged in slow, hands-on processes of data annotation, model training, and interpretation, fostering critical reflection on their creative practices and technical understanding of AI. The study highlights the potential of material and temporal disentanglement to cultivate agency and reflective practice in design education. Future work could extend this approach to generative AI systems and professional design contexts, offering a meaningful alternative to efficiency-focused AI tools.