Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

ReflectEd: Evaluating Reflection-Driven Learning in an AI-Assisted System

Md Nazmus Sakib, Ishika Tarin, Naga Manogna Rayasam, Manas Gaur, Sanorita Dey

Abstract

In collaborative settings, sustaining momentum and engagement between checkpoints (e.g., meetings) can be challenging, often leading to task drift and reduced preparedness. To address this gap, we developed ReflectEd, an AI-assisted system that supports between-checkpoint reflection through theory-driven prompts with progressively structured levels and mechanism-based scaffolding. We evaluated ReflectEd in a mixed-method study comparing two reflection configurations: a regular reflection workflow and a deeper reflection workflow that included an additional transformative reflection activity. Across conditions, participants reported steady engagement early in the week. In the deeper configuration, later reflections tended to exhibit higher actionability and richer forward-looking planning, while also being harder to sustain and more effortful during periods of active work. Partner-visible reflections were frequently described as supporting coordination by surfacing differences in focus and facilitating accountability. Overall, the findings characterize trade-offs between reflection depth, feasibility, and perceived preparedness for subsequent checkpoints. We discuss implications for the design of AI-assisted systems that support collaboration readiness and reflection-oriented regulation in time-constrained collaborative workflows.

ReflectEd: Evaluating Reflection-Driven Learning in an AI-Assisted System

Abstract

In collaborative settings, sustaining momentum and engagement between checkpoints (e.g., meetings) can be challenging, often leading to task drift and reduced preparedness. To address this gap, we developed ReflectEd, an AI-assisted system that supports between-checkpoint reflection through theory-driven prompts with progressively structured levels and mechanism-based scaffolding. We evaluated ReflectEd in a mixed-method study comparing two reflection configurations: a regular reflection workflow and a deeper reflection workflow that included an additional transformative reflection activity. Across conditions, participants reported steady engagement early in the week. In the deeper configuration, later reflections tended to exhibit higher actionability and richer forward-looking planning, while also being harder to sustain and more effortful during periods of active work. Partner-visible reflections were frequently described as supporting coordination by surfacing differences in focus and facilitating accountability. Overall, the findings characterize trade-offs between reflection depth, feasibility, and perceived preparedness for subsequent checkpoints. We discuss implications for the design of AI-assisted systems that support collaboration readiness and reflection-oriented regulation in time-constrained collaborative workflows.
Paper Structure (14 sections, 3 figures, 2 tables)

This paper contains 14 sections, 3 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Related Works
  3. Learning Through Reflection and Reflection-on-Action
  4. Intelligent Systems and Human-Centered AI in Collaborative Work
  5. Methodology
  6. Theoretical Baseline of ReflectEd:
  7. Findings
  8. Quantitative Analysis
  9. Reflection Progression:
  10. Qualitative Analysis
  11. Discussion
  12. Design Tradeoffs in Reflection Depth:
  13. Reflection as a Learning Scaffold:
  14. Limitations and Future Work:

Figures (3)

  • Figure 1: Overview of the multi-day reflection workflow. Each checkpoint meeting is summarized by GPT-4o-mini and incrementally enriches a shared context history across descriptive, explanatory, relational, and transformative reflection stages. Note: Day 6 reflection is only applicable in the Deeper condition.
  • Figure 2: Example snapshots from ReflectEd for Day 3, Day 5, and Day 6 reflection activities.
  • Figure 3: Reflection and actionability progression across days by condition.