Upper School AI Use & Integrity Policy

Whether or not teachers use AI in class, students will encounter it outside of school. Our goal is to teach responsible, transparent, and ethical use of AI, while maintaining academic integrity and supporting skill development. Mistakes can lead to reteaching, while repeated or intentional misuse leads to consequences. 

FOUNDATIONAL EXPECTATIONS (APPLIES TO ALL CLASSES)

1. Transparency and Clarity 

By default, AI is not permitted on assignments. If AI is allowed, teachers will clearly indicate one of the following: 

• AI Allowed for Specific Tasks (listed) 
• AI Allowed with a Required “How I Used AI” Note 

Students should assume AI is not allowed unless a teacher explicitly states otherwise. 2. Process Expectations (Protect Students & Teachers) 

To support fairness, learning, and integrity: 

• Major writing assignments must be drafted in Google Docs (or another teacher-approved platform) so the writing process is visible through the version history 
• Students should not delete drafts or any documents that showcase progress until the assignment or unit is complete 
• If a plagiarism/AI detector raises concerns, teachers will speak with the student first to review their process. To better understand a student’s authorship and learning process, teachers may also ask the students to provide a brief handwritten summary explanation, or oral response related to the assignment. These checks focus on understanding, not punishment, and help to ensure the work reflects their own thinking. 
• Automated detectors are used as conversation starters, not evidence of wrongdoing. ○ Some checkers that may be used are: 

■ GPTZero: Detects text likely generated by large-language models. Provides sentence-level AI probability and “human vs AI” scoring. 

■ Copyleaks: Combines AI detection with plagiarism detection, offers detailed reports, and handles essays and research-type writing well. 

■ Scribbr AI Detector: Offers human vs AI detection, flags content generated or heavily refined by AI, works across multiple languages. 

■ Winston AI: Built for educators and provides visual “human vs AI” breakdowns, supports bulk uploads, integrates with LMS, and claims high accuracy; useful for large classes or frequent submissions. 

○ It should be clearly noted that though AI detectors are helpful, they are not foolproof and give False-Positives and/or False-Negatives. So the use of Google Docs (for example) where an instructor can view the students editing process and progress is critically important. 

■ AI Detection tools are not able to reliably identify all AI-generated content, especially when text has been heavily revised. For this reason, the writing process, version history, and student explanation are more important than detection results alone. 

■ A flagged sentence, or a few, does not mean the work is AI created, or demonstrates academic dishonesty.Academic dishonesty occurs when AI is used in a way that replaces a student’s own thinking, planning, or expression on an assignment that is meant to reflect their own independent work. 

• Mistakes lead to reteaching. Repeated choices lead to consequences. 
• Teachers may reduce credit within the suggested range, based on student effort, honesty, and quality of revised work at their discretion. 

1. What Counts as AI? (Student-Friendly Definitions) 

Allowed (unless teacher says otherwise): 

• Spellcheck/grammar check 
• Formatting tools 

Requires teacher permission: 

• Sentence rewriting or rephrasing 
• Translation tools 
• Paragraph-level revision suggestions 
• Generating images, charts, or diagrams 
• Generating ideas, concepts, characters, or themes 

Not allowed unless explicitly assigned: 

• Generating sentences, paragraphs, essays, or answers 
• AI problem solvers for math, science, or coding 
• Paraphrasing to hide copied material 
• Using AI to misrepresent one’s thinking 

Important Note: The planning, brainstorming, and conceptualization stages of an assignment are a critical part of the learning process. Unless a teacher explicitly allows it, students should not use AI to generate ideas, themes, characters, or project concepts. Students may request permission for limited AI use for specific tasks within an assignment. 

 

GUIDELINES: 

• Indicators of Authenticity 

  • Work shows conceptual understanding 

  • Document history supports student authorship 

  • Student can explain ideas and revisions 

• Indicators of Concern 

  • Student denies AI use despite evidence 

  • Entire sections appear suddenly in final draft 

  • No meaningful revision history 

 

TIERED CONSEQUENCE SYSTEM FOR AI MISUSE 

How Consequences Are Determined 

Consequences are determined by the severity of the misuse, the student’s intent, and whether a pattern of behavior is present. This means that repeated misuse in different classes is considered part of the same pattern and may result in escalated consequences. 

Tier	Intent	Pattern	Impact	Involvement
1	Mistake	None	None	Parent notification is optional
2	Poor Judgement	Emerging	Partial	Parents emailed, student notes updated
3	Dishonesty	Clear	Zero+Recover y	Parents and Admin are emailed
4	Severe	Repeated	Restrictive  Plan	Parents and Admin are emailed

TIER 1: FIRST INSTANCE (LOW-LEVEL, ACCIDENTAL MISUSE, MISTAKE, OR CONFUSION) 

Examples 

• Student uses Grammarly, ChatGPT, etc. to “improve wording” when not allowed ● Student uses a translator instead of writing in English or foreign language ● Small portions of AI-generated text appear in the work 
• Student seems confused rather than malicious 

Response 

• Restore & Repair: Student revises or redoes the work without AI (or with proper disclosure) 
• Teacher–Student Conversation (short, documented) 
• Light parent notification (email) 
• Mini-lesson: Teacher reviews class AI expectations 
• No academic penalty if misunderstanding appears genuine 

Goal: Clarify expectations early, prevent shame, and build responsible habits. 

 

TIER 2: SECOND INSTANCE (PATTERN OR DISHONESTY) 

Examples 

• Student uses AI to write significant parts of an assignment 
• Multi-class pattern emerges 
• Student denies AI use when version history or process evidence suggests otherwise ● AI used on independent work or homework meant to show their thinking 

Response 

• Redo assignment with teacher guidelines 
• Reduced credit on redo (typical range: –10% to –25%) 
• Parent email 
• Student notes updated 

Goal: Address pattern behavior, reinforce honesty, support skill-building. 38

TIER 3: THIRD INSTANCE (MAJOR INTEGRITY VIOLATION) 

Examples 

• Entire assignment or major portion is AI-generated 
• AI used on a quiz or assessment 
• Evidence of intentional misrepresentation or evading teacher expectations ● Repeated offenses despite support and reteaching (from Tiers 1 and 2) 

Response 

• Assignment is scored as 50%, with the option to complete a recovery assignment for partial credit 
• Mandatory Family Meeting (student, parent, teacher, and Upper School Director) 
• Potential loss of Chromebook/laptop will be evaluated given the contributing factors behind the AI misuse 

Goal: Reset the pattern, restore trust, and reinforce academic integrity. 

 

TIER 4: SEVERE OR REPEATED VIOLATIONS 

Examples 

• Using AI to harass, impersonate, or generate harmful content 
• Repeated academic dishonesty after Tier 3 interventions 
• Circumventing school filters, using unauthorized accounts, or bypassing restrictions ● Coordinated academic dishonesty involving AI or shared AI outputs Response 
• Assignment is scored as a 50% with no opportunity to make it up 
• Mandatory Family Meeting (student, parent, teacher, Upper School Director, and Head of School) 

○ Evaluate the situation to determine whether an at-home reflection period is needed 

• Quarter-long AI-restricted plan will explore: 

○ Hand-written assignments 

○ Loss of Chromebook/laptop privileges 

Goal: Protect the community, uphold integrity, and ensure digital safety.