Plan: Expectations, Boundaries, & Disclosures

A course syllabus is your contract of trust. CSS has adopted an AI Ethical Use Statement and drafted an optional Syllabus Statement that faculty should place in the syllabus in addition to an explanation of acceptable, appropriate, and/or required uses of AI (and why this is so). Why? How we evaluate student work dictates how students engage with technology. The AI Assessment Scale (AIAS) enables us to specify different levels of permitted technology use based on what we are actually measuring, which can reduce student confusion. 

If you'd like examples, Yost (2024) provides several options: AI Syllabus Statements and Course Policies. Also see The AI Assessment Scale to explore each level below further or try their Custom GPT to guide you through implementing this framework.

Level 1: No AI (Completely Restricted)

• The Cognitive Focus: Foundational knowledge retention, spontaneous critical thinking, and authentic, unassisted personal voice. The student executes 100% of the creative and cognitive work. AI tools are completely restricted.

• Concrete Assessment Examples:

  • The Live Socratic Viva Voce: A 5-minute, one-on-one or small-group live oral discussion where students verbally explain their understanding of a core course concept or case study.

  • In-Class Structural Mapping: A completely unassisted, handwritten concept-mapping or blueprinting exercise completed entirely during class time on paper or a monitored local screen.

  • Immediate Blue-Book Reflection: A timed, in-class response to a sudden, highly localized prompt (e.g., analyzing a specific guest speaker's presentation from the night before) that leaves no window for external generation.

Level 2: AI Planning (Brainstorming & Ideation)

• The Cognitive Focus: Conquering "blank page anxiety," project management, and organizing complex thoughts. The AI acts purely as a sounding board or structural outliner, but the student writes every word of the final submission from scratch.

• Concrete Assessment Examples:

  • The Flipped Outline: Students prompt an AI to generate three distinctly different structural outlines for an essay topic. They turn in a brief critique explaining which outline is best and why, and then write the actual paper manually.

  • The Topic Sandbox: Students use a chatbot to brainstorm five narrow research questions stemming from a broad theme, vet them for feasibility against library databases, and submit their final chosen topic with the AI conversation log attached.

  • Project Milestones & Gantt Charting: For long-term group projects, students use AI to break down a massive rubric into a weekly project-management schedule, assigning roles and deadlines, which they then follow to execute the work manually.

Level 3: AI Collaboration (Iterative Co-Editing)

• The Cognitive Focus: Line-editing, mechanical refinement, stylistic improvement, and argumentative polishing. The student directs the workflow, provides the core thesis, and makes all final executive decisions; the AI acts as a collaborative editor or technical reviewer.

• Concrete Assessment Examples:

  • The "Tracked Changes" Revision Log: Students write a complete first draft independently, paste it into an LLM with a specific editing prompt, and submit their final polished paper alongside an appendix detailing exactly which AI edits they accepted, which they rejected, and why.

  • The Reverse-Role Writing Coach: Students feed their rough draft to an AI with the prompt: "Act as a critical reviewer. Do not rewrite my text, but point out three weak transitions and two logical gaps." Students turn in the AI's critique alongside their revised, strengthened draft.

  • Bilingual Stylistic Adaptation: A non-native English speaker writes an assignment to demonstrate their content knowledge, then pairs with an LLM to refine the sentence flow, submitting the raw original text alongside the polished version to demonstrate mechanical growth.

Level 4: Full AI (Human Curation & Quality Assurance)

• The Cognitive Focus: Prompt engineering, fact-checking, systematic auditing, and high-level evaluation. The AI generates the bulk of the raw content or processes complex datasets, while the student acts as the managing editor and quality control expert.

• Concrete Assessment Examples:

  • The AI Performance Review: Students use an LLM to generate a comprehensive business case study solution or a computer code script. The student then acts as a senior supervisor, marking up the machine's output with redlines to correct hallucinations, superficial analysis, or generic formatting.

  • The Comparative Synthesis: Students prompt two different AI models (e.g., ChatGPT vs. Gemini) using identical, sophisticated prompts (like COSTAR) to solve a comple/x multi-step problem. They write an analytical critique comparing the depth, accuracy, and biases of both outputs.

  • The Prompt Chronology Portfolio: Instead of grading a final product, the instructor grades a detailed "Prompt Portfolio." Students turn in a chronological diary of a long, multi-turn AI conversation showing how they continuously adjusted their personas, constraints, and context to force the AI to yield a highly specific, professional result.

Level 5: AI Exploration (Tool Evaluation & Literacy)

• The Cognitive Focus: Metacognitive awareness, digital citizenship, and technical literacy. Students explicitly experiment with the changing nature of AI technology itself, analyzing the behavior, algorithmic boundaries, and ethical implications of specialized tools.

• Concrete Assessment Examples:

  • The Bias & Inclusivity Audit: Students prompt an image or text generator to create profiles of professionals in a specific industry (e.g., "successful CEOs" or "software engineers") and write a critical reflection analyzing the systemic societal stereotypes and data biases replicated in the output.

  • The Hallucination Deep Dive: Students prompt a general chatbot to conduct a niche academic literature review. They must track down every single citation provided by the AI in the real-world university library databases to verify its legitimacy, cataloging exactly which sources were real and which were entirely fabricated by the machine.

  • The Workflow Efficiency Experiment: Students complete a complex academic task twice—once entirely manually, and once using specialized research agents (like Elicit or Consensus). They submit a comparative analysis tracking the time spent, the depth of learning achieved, and the ethical trade-offs regarding data privacy and environmental impact.

Academic Integrity

When teachers attempt to deter cheating with severe penalties, such as failing the course, the burden of proof is on the accuser. AI use is challenging to detect because detection tools are not 100% accurate, particularly when grammar tools are used. 

That doesn't mean, however, that nothing can be done to prevent cheating. Teaching Students About Plagiarism (2:44) or teaching them how to use AI ethically (2026 Student Guide to AI from Elon University and the AACU) can reduce or even prevent cheating. 

Here are some definitions to assist you (from: Frontier, T. (2025, May 11). Catch Them Learning: A Pathway to Academic Integrity in the Age of AI. Cult of Pedagogy.).

• Cheating is using a tool or resource to misrepresent one’s knowledge and skills to receive undue credit for a task. The accuser needs to support a claim about cheating with evidence that cheating occurred. 

• Integrity describes a commitment to ensure one’s completion of a task accurately represents the knowledge and skills one actually possesses. A claim about integrity is supported by a learner’s ability to transparently document, and explain the learning process and their results.

• Transparency occurs when students document their steps and resources used to prepare for and engage in a task.

• Explainability occurs when students can explain, expand, and reflect on what they’ve learned and their evidence of learning.

For more options, see AI Syllabus Statements and Course Policies by Yost, L. (2024) and Chapter 4, Setting Policy and Communicating Expectations by Heaps, T. (2024).  

Managing Suspected Misconduct

Since AI detectors are unreliable and biased against neurodivergent and non-native English speakers, include a process review meeting clause in your syllabus:

"If the teaching team has questions regarding the authorship of a submitted assignment, we reserve the right to pause grading and invite you to an informal process review meeting to discuss your work process, source materials, and the conceptual ideas behind your work."

This shifts the encounter from a punitive accusation into a productive conversation.

Misuse Decision Tree

Is there a suspected AI violation?

1. Document the Evidence 

   • Compare the submission to known student writing benchmarks. 

   • Note anomalies (e.g., hallucinated citations, generic tone). 

2. Bypass Automated Detectors

   • Remind yourself: Detectors have high false-positive rates and systemic biases against diverse learners. 

3. Trigger the Socratic Dialogic Review Clause

   • Invite the student to a supportive, low-stakes chat.

   • Ask: "Walk me through your research process."

   • Ask: "Can you explain the core concept behind this paragraph?"

4. Human-Centered Resolution

   • If the student understands the concepts: Treat as a learning success. 

   • If the student cannot explain the work: Pivot to a developmental rewrite or follow standard CSS academic integrity procedures. 

Why Disclose

Disclosing AI use to students is one of the most powerful, yet overlooked, strategies for faculty to build a transparent and equitable learning environment. Here is why faculty disclosure matters:

1. It Models Academic Integrity. We cannot expect students to be transparent about their AI use if we hide our own. Openly disclosing your use demonstrates honest digital citizenship in action.

2. It Shows Students Where the "Line" Is. Students are anxious about what constitutes cheating. When you model a clear disclosure framework in your own work, you give them a concrete, practical blueprint of what responsible use looks like.

3. It Levels the Playing Field. Tech-savvy students easily spot AI-generated assignments and rubrics, while others do not. Disclosing your use ensures all students have the same baseline understanding of how their course materials were made.

4. It Builds Classroom Trust. Education is now a three-way interaction between the teacher, the student, and the technology. Being honest about how your course was constructed builds deep foundational trust, making students much more willing to respect boundaries when you ask them to lock down the AI for an unassisted assignment.

When & How to Disclose 

Here are sample disclosures based on the level of your AI use.

1. Level 1: Minimal

   • Scope: Purely brainstorming, suggesting prompt variations, or refining pre-written paragraphs. AI does not shape pedagogical intent.

   • Disclosure: None required in final materials.

2. Level 2: Moderate

   • Scope: Generating specific content elements (e.g., discussion prompts or draft outlines). Educator revises and integrates.

   • Disclosure: Clear disclosure to learners in the syllabus or LMS.

3. Level 3: Significant

   • Scope: AI forms the actual backbone/structure of major lessons or assessments. Humans focus on validation and curation.

   • Disclosure: Note when AI-generated material comprises a visible part of the course.

4. Level 4: Comprehensive

   • Scope: AI is deeply integrated; producing the vast majority of course content, media, or adaptive tracks.

   • Disclosure: Explicit context for learners regarding system limits and human validation measures.