Assessment and Academic Integrity

Traditional assessment practices assumed students would complete work independently without sophisticated assistance. AI fundamentally challenges this assumption. This week addresses how to rethink assessment while maintaining academic integrity.

The assessment challenge

What AI changes

AI can produce competent responses to most traditional assessments:

• Essays and written responses
• Problem solutions with explanations
• Research summaries
• Creative writing
• Code and technical solutions

This creates a validity problem: Do assessments measure student capability or AI capability?

What we actually want to assess

Knowledge: What students know and understand Skills: What students can do Thinking: How students reason and problem-solve Growth: How students have developed over time

Traditional assessments often measured proxies for these. AI forces us to measure more directly.

The detection illusion

Many educators hope detection technology will solve the problem:

• AI detection tools are unreliable
• Students can easily evade detection
• False accusations harm innocent students
• Detection creates adversarial relationships

A better approach: Design assessments that make AI use irrelevant or visible.

AI-resistant assessment design

Principle 1: Assess the process, not just the product

When you can see how work was done, AI assistance becomes visible.

Strategies:

• Require drafts and revisions with timestamps
• Include in-class components
• Use oral examinations or presentations
• Document the thinking process

Principle 2: Personalize the task

Generic prompts are easy for AI. Personal contexts are harder.

Strategies:

• Connect to specific class discussions
• Reference individual student experiences
• Build on previous student work
• Include locally-specific content

Principle 3: Emphasize original thinking

AI is good at synthesis; original thought is harder.

Strategies:

• Ask for novel applications of concepts
• Require connections to personal experience
• Demand original analysis, not summary
• Focus on judgment and evaluation

Principle 4: Make AI use transparent

When AI use is disclosed and appropriate, integrity is maintained.

Strategies:

• Allow AI for specified portions of work
• Require reflection on AI contributions
• Assess the human additions to AI work
• Evaluate AI use choices as part of the grade

Types of AI-resistant assessments

In-class assessments

Work completed in controlled settings:

• Timed essays or problems
• Open-note examinations
• Group problem-solving sessions
• Lab practical assessments

Benefits: Clear verification that work is student-produced. Limitations: Not all learning shows in timed conditions.

Oral assessments

Verbal demonstration of understanding:

• One-on-one conferences
• Presentation with questions
• Defense of written work
• Socratic discussion

Benefits: Direct assessment of student thinking. Limitations: Time-intensive for large classes.

Performance assessments

Demonstration of skills in context:

• Laboratory investigations
• Physical demonstrations
• Real-time problem-solving
• Collaborative projects with individual accountability

Benefits: Authentic measure of capability. Limitations: Requires careful design and observation.

Portfolio assessments

Collection of work over time with reflection:

• Evidence of growth
• Process documentation
• Self-assessment components
• Conference or presentation

Benefits: Holistic view of learning. Limitations: Requires ongoing documentation.

Process-based assessment

Documenting the journey

Require evidence of process alongside final products:

• Brainstorming records
• Research notes with sources
• Draft progression
• Revision reflections

This makes AI assistance visible without prohibiting it.

Metacognitive reflection

Ask students to reflect on their learning:

• What was challenging and why
• How understanding developed
• What strategies were used
• How AI was or was not used

Genuine reflection is difficult to fabricate.

Conference-based assessment

Discuss work with students:

• Verify understanding of their own work
• Explore reasoning and choices
• Address questions about AI use
• Provide personalized feedback

Conversations reveal what students actually know.

Balancing assessment types

AI-restricted assessments

For measuring fundamental skills and knowledge:

• Basic skill verification
• Core content understanding
• Timed and controlled conditions

Purpose: Ensure foundational competency.

AI-allowed assessments

For measuring higher-order skills:

• Complex problem-solving
• Research and synthesis
• Creative production
• Professional-style work

Purpose: Develop AI-collaborative skills.

Transparent about purpose

Communicate why different assessments have different rules:

• Skill verification requires independent demonstration
• Complex projects benefit from AI collaboration
• Both are legitimate learning goals

Maintaining integrity culture

Beyond rules

Culture matters more than rules:

• Model integrity in your own practice
• Discuss why integrity matters
• Recognize honest effort
• Address violations with education, not just punishment

Student understanding

Help students understand:

• Why certain assessments restrict AI
• What integrity means in AI context
• How violations affect their learning
• The long-term value of genuine skill development

Peer influence

Create positive peer pressure:

• Celebrate academic integrity
• Make integrity part of community identity
• Enable peer accountability without surveillance

Technology and assessment

Digital tools for integrity

Technology can support integrity:

• Revision history tracking
• Time-stamped documentation
• Controlled assessment environments
• Process recording tools

Appropriate use of proctoring

If proctoring is used:

• Explain the purpose clearly
• Minimize invasiveness
• Use for high-stakes assessments only
• Consider equity implications

Balance technology and trust

Technology should support culture, not replace it:

• Over-surveillance damages trust
• Students respond to expectations
• Technology cannot substitute for relationship

Redesign process

Step 1: Audit current assessments

Review existing assessments:

• What do they actually measure?
• How vulnerable are they to AI?
• What is the purpose of each?

Step 2: Identify priority changes

Focus on highest-stakes and most vulnerable assessments first.

Step 3: Design alternatives

Use principles above to redesign assessments that maintain validity.

Step 4: Communicate changes

Explain new approaches to students, parents, and colleagues.

Step 5: Iterate and improve

Gather feedback and refine assessment approaches over time.

Common assessment mistakes

Mistake 1: Relying on detection

Detection-based strategies are fundamentally flawed. Design around the problem.

Mistake 2: Eliminating all outside-class work

Some learning requires extended work. Adapt rather than abandon.

Mistake 3: All assessments treated the same

Different purposes require different approaches.

Mistake 4: No communication about changes

Students need to understand why assessments are designed as they are.

Mistake 5: Ignoring equity

Assessment changes should not disadvantage students with fewer resources.

Key takeaway

AI forces us to clarify what we actually want to assess and design assessments that measure it directly. Focus on process and verification rather than detection. Balance AI-restricted and AI-allowed assessments based on learning objectives. Build a culture of integrity that goes beyond rules to genuine understanding of why integrity matters.