Bridging Integrity and Innovation: An AI Agent to Monitor, Evaluate, and Validate Student Effort in AI-Assisted Coursework

 

WIUT AI-Assisted Learning Effort Agent Proposal

Bridging Integrity and Innovation: An AI Agent to Monitor, Evaluate, and Validate Student Effort in AI-Assisted Coursework


2. Executive Summary

With the rapid adoption of generative AI tools like ChatGPT, higher education institutions face a new challenge: distinguishing authentic learning from AI-generated outputs. This proposal introduces an AI Agent that acts as an intelligent intermediary between students and AI tools. It records prompts, evaluates learning interactions, and calculates an "Effort Score"—creating a transparent learning trace. Once a student reaches a predefined effort threshold, their use of AI is considered constructive and permissible, supporting authentic learning outcomes and academic integrity.


3. Problem Statement

  • Students increasingly use AI tools to generate coursework.
  • Academics are burdened with detecting AI usage, often using questionable detection tools.
  • A combative environment undermines trust, and shifts focus away from learning.
  • Genuine learners who engage with AI meaningfully are unfairly penalized.

4. Solution Overview: The AI Effort Evaluation Agent

What it does:

  • Acts as a proxy between the student and AI tools (e.g., ChatGPT).
  • Logs every prompt and response.
  • Evaluates effort in real time using NLP heuristics and metadata.
  • Builds a dynamic Effort Score that only increases.
  • Generates a final report that includes:
    • Full interaction history
    • Learning summary
    • Effort milestones
    • AI transparency log

Outcome:

  • Student may submit AI-assisted work with full transparency.
  • Academics see a traceable learning journey.
  • Integrity is preserved, learning is encouraged.

5. Flowchart: AI Agent Interaction Lifecycle

Student Login   ->  Prompt Sent to AI Agent -> Agent Forwards to ChatGPT    <- Agent Receives & Logs Response 

         ---> Calculate Effort Rating

         ---> Analyze Prompt Depth

         ---> Detect Iterative Refinement

Update Cumulative Effort Score 

Generate Learning Summary 

Threshold Reached?       |

|   (e.g., Score >= 100)   |

--------------------------

        /      \

      Yes      No

      /          \

--------------------- ----------------------

| Allow Final Export  |   | Keep Logging Efforts |

| With Auth Summary  


6. Core Components & Architecture

1. Frontend (Student Interface)

  • Chat-like UI
  • Visual tracker of Effort Score
  • Request final report

2. Backend Agent Server

  • Node.js / ASP.NET Core API
  • Prompt logging
  • Effort calculation engine
  • Learning summary generator

3. Effort Calculation Engine (Core IP)

  • Prompt quality heuristic
  • Interaction richness (follow-ups, clarifications)
  • Meta-learning tracking (e.g., summary requests, code iterations)
  • No penalty, only reward (effort only grows)

4. Database (PostgreSQL / MongoDB)

  • User sessions
  • Prompt–response pairs
  • Timestamped effort logs
  • Exportable JSON / PDF reports

5. Educator Dashboard

  • Visual trace of AI usage per student
  • Report download
  • Optional feedback interface

7. Benefits to WIUT

Reduces academic dishonesty stress
Encourages constructive use of AI
Provides evidence-based evaluation
Scalable to UG, PG, and PhD levels
Reinforces WIUT's commitment to innovation with integrity


8. Future Enhancements

  • Plagiarism + Effort Correlation Reports
  • Integration with  SRS
  • Peer review option
  • Feedback sentiment analysis
  • AI tutor suggestions based on student effort areas

9. Call to Action

We propose to:

  • Build a working prototype in 2–3 months
  • Pilot with select students and academic supervisors
  • Present dashboard and interaction reports to validate feasibility

10. Appendix: UI Mockups (To Be Designed)

  • Student chat screen
  • Live effort tracker
  • Educator dashboard
  • Final summary PDF

 https://chatgpt.com/share/68344037-efc4-800c-98ca-cba86a36ec45 

Comments