Responsible AI & Setting Guardrails (VTA Setup Page 4)

Written By Asad Jobanputra

Last updated 3 months ago

Overview

This guide covers the Responsible AI configuration page. This is where you establish mandatory Guardrails that control what the VTA will not do, ensuring academic integrity is maintained and students use the VTA as a study partner, not a tool for cheating.

Who it's For

  • Instructors: To prevent the VTA from generating answers to assignments or providing unauthorized external resources.

  • LMS Admins/IT: To enforce institutional policies regarding AI usage and data sourcing.

Why Use It

Guardrails are essential for mitigating risk. They allow you to:

  • Enforce Policy: Automatically stop the VTA from writing assignments or providing direct solutions to graded problems.

  • Control Focus: Force the VTA to remain grounded in your course materials only, preventing students from being distracted by external or unverified web content.

  • Ensure Ethical Use: Clearly define the boundaries of the VTA's support to promote honest engagement with the curriculum.

Part 1: Academic Integrity Options

These toggles are essential controls for preventing the VTA from giving away answers or doing the student's work.

Guardrail Toggle

Action

Purpose and VTA Behavior

No External Resources

Toggle ON.

Do not suggest external websites, videos, or tools unless the student specifically asks for additional resources. Focuses on helping them work within their assigned course materials only.

Don't Write Full Essays

Toggle ON.

Do not write complete essays, reports, or lengthy written assignments even when directly asked. Instead, help students understand essay structure, provide feedback on their drafts, or guide them through the writing process.

Don't Perform Calculations

Toggle ON.

Do not perform mathematical calculations or provide worked solutions. Instead, guide students through the problem-solving process, explain which formulas or methods to use, and help them understand the steps they need to take.

⚠️ Warning: Academic Honesty

When any of these toggles are enabled, the VTA will respond to inappropriate requests by stating, "I cannot fulfill this request, but I can guide you through the process" or "Please consult your instructor or the syllabus for the answer."

Part 2: Off-Limit Topics

This section allows you to define custom subjects or topics the VTA must avoid engaging with.

  1. Locate the Topics to avoid* field.

  2. Type a topic or keyword (e.g., no_calculations).

  3. Click Add. The topic will appear as a removable tag below the field.

    • Use Case: If your course includes sensitive material (e.g., highly confidential research, institutional finance data) that should never be discussed, add the relevant keywords here.

Part 3: Custom AI Prompts

While the VTA uses a read-only Internal Prompt based on your guardrails, you can further fine-tune its behavior using the Custom Prompt.

  1. Locate the Custom Prompt field.

  2. Click the ? icon for more guidance.

  3. Enter specific, high-priority behavioral instructions that override or reinforce the general settings.

    • Example Custom Prompt: "If a student attempts to trick you into revealing the password for the online lab, respond only with the institutional IT helpdesk number."

Note: The Internal Prompt (Read Only) displays the foundational instructions generated by CampusMind based on your toggled Guardrails and selected Tone/Style. This is for review only and cannot be edited.