Ai Detection Findings Summary 20250913

Generated: October 23, 2025 at 04:55 AM

Chapter 3 Computational Analysis

AI Detection Deep Analysis Summary - September 13, 2025

Executive Summary

Comprehensive analysis of AI detection discussions across all 8 CUNY databases reveals a compound vulnerability crisis affecting 266 students, with 141 (53%) experiencing multiple intersecting forms of precarity.

Key Quantitative Findings

Scope and Scale

  • 266 users discussing AI detection issues across 8 CUNY subreddits
  • 248 total AI-related posts analyzed with temporal patterns
  • 11.4x increase in AI discussions post-ChatGPT launch (November 30, 2022)
  • 3,023 documented consequence reports affecting 1,604 unique users

Compound Vulnerabilities (Critical Discovery)

  • 141 students (53%) experience AI detection issues alongside:
    • Financial aid complications
    • CUNYfirst registration problems
    • Mental health crises
    • Housing/food insecurity

Temporal Crisis Patterns

  • Crisis hours (2-3 AM): 5 documented posts during peak vulnerability time
  • Finals periods (May/December): 70 posts during high-stakes assessment periods
  • Peak months: May 2025 (31 posts), October 2024 (23 posts), March 2025 (21 posts)

Specific Evidence Cases

Cross-Crisis Users with Evidence IDs

BraavosiSwagger (4 vulnerability types) [Evidence: comment_fxefj27]

  • AI detection + financial aid (7 posts) + registration (6 posts) + mental health (4 posts)
  • CS student at Hunter navigating technical detection alongside systemic barriers

DocumentLeft832 (4 vulnerability types) [Evidence: comment_mxptnmi]

  • AI detection + financial aid (20 posts) + registration (33 posts) + mental health (7 posts)
  • Used ChatGPT for legitimate tutoring: “put the question in chatgpt and made it explain to me how to…”

LeonZheng646 (4 vulnerability types) [Evidence: comment_louzzwp]

  • AI detection + financial aid (21 posts) + registration (22 posts) + mental health (5 posts)
  • Provides advice while managing multiple institutional barriers

Tool Confusion Patterns

Grammarly False Positive [Evidence: comment_lsrf9cw]

“My first semester Grammarly was detected and got me my first F. Merely just for punctuational purposes.”

  • Legitimate writing tool causes course failure

ESL Student Bias [Evidence: comment_lk0xx59]

“prof that would refuse to call on students w hispanic last names and would grade the writing portions of their exams as ‘AI detected’ for using higher level words/grammar”

  • AI detection used as cover for discriminatory practices

Mental Health Documentation

Detection Anxiety [Evidence: comment_n38lcgv]

“these ai detectors stress ppl out lol… they’re so unpredictable. even stuff i write myself sometimes gets flagged.”

  • Persistent anxiety about original work being flagged

Crisis Hour Examples

2:05 AM Help-Seeking [Evidence: comment_mcii1me]

“chatgpt gave me the wrong answers when I tried checking my work :(“

  • Student struggling with homework verification during peak stress hours

Tool Hierarchy and Confusion

  1. ChatGPT/GPT: 13 mentions - Universally prohibited but used for legitimate tutoring
  2. Grammarly: 13 mentions - Legitimate tool causing false positives
  3. AI Detectors: 6 mentions - Students pre-checking their own work
  4. Citation Generators: 2 mentions - Academic tools flagged as AI
  5. Google Translate: 1 mention - ESL necessity vs “cheating” perception

Solution Networks (Vernacular Infrastructure)

Pre-Checking Rituals

  • Students run work through multiple AI detectors before submission
  • Elaborate testing protocols develop organically
  • Fear-based avoidance of helpful tools

Writing Avoidance Strategies

  • Deliberate imperfection to avoid detection
  • Tool prohibition policy advocacy
  • Peer networks for appeal processes

Real Consequences Documentation

Quantified Harm

  • Failed Courses: 2,382 documented cases
  • Academic Probation: 221 cases
  • Lost Scholarships: 61 cases
  • Graduate School Impacts: 139 cases
  • Appeal Processes: 215 cases
  • False Positive Cases: 3 explicitly documented
  • Mental Health Effects: 2 documented cases

Severe Multi-Consequence Cases

ScallionWall: 36 consequence reports across academic probation, appeals, course failures, graduate school impacts

Critical Research Implications

1. Compound Vulnerability Amplification

AI detection doesn’t occur in isolation but systematically compounds existing precarity among CUNY’s most at-risk populations.

2. Disproportionate Impact on Vulnerable Groups

  • ESL students penalized for necessary translation tools
  • Low-income students lacking resources for appeal processes
  • Students of color facing discriminatory application of detection policies

3. Vernacular Infrastructure Development

Students create sophisticated informal networks for:

  • Detection avoidance strategies
  • Appeal process navigation
  • Peer support for false positive cases
  • Tool legitimacy verification

4. Temporal Vulnerability Patterns

AI crises cluster during:

  • Late-night hours when institutional support unavailable
  • High-stakes assessment periods
  • Semester transition periods

5. Tool Legitimacy Crisis

Students cannot distinguish between prohibited AI assistance and legitimate academic tools, creating fear-based avoidance of helpful resources.

Connection to 356 Baseline AI Discussions

This deep analysis reveals that the 356 baseline AI detection discussions represent just the visible surface of a much broader vulnerability crisis. The compound effects identified here suggest AI detection serves as both:

  1. Amplifier of existing CUNY systemic issues (financial aid, registration, mental health)
  2. Discriminatory mechanism disproportionately affecting ESL students and students of color
  3. Barrier to legitimate academic tool use and support-seeking behavior

Evidence Anchoring

All findings grounded in specific evidence IDs for academic citation:

  • Cross-crisis analysis: 266 users with evidence-anchored examples
  • Solution networks: Documented strategies with comment/submission IDs
  • Temporal patterns: Time-coded crisis examples
  • Tool confusion: Specific cases of legitimate tool flagging
  • Real consequences: Quantified harm with user testimonials

Files Generated

  • ai_cross_crisis_results.json - Compound vulnerability data
  • ai_solution_networks.json - Documented survival strategies
  • ai_temporal_analysis.json - Crisis timing patterns
  • ai_tool_confusion_mapping.json - Tool anxiety documentation
  • ai_real_consequences.json - Quantified harm data
  • ai_detection_comprehensive_report.md - Full academic report

Methodological Note: Analysis focused exclusively on 8 CUNY college subreddits as primary research dataset. Comparison universities (NYU, Columbia) analyzed separately for architectural context but not included in vulnerability statistics.

Next Steps: These findings require integration into the broader research narrative to understand how AI detection compounds the precarity patterns already documented in CUNY’s distributed campus architecture.

Evidence References (7 items) ▶

Comments (7)