Lockdown browsers for online test proctoring have been found to falsely accuse black users for cheating 34-35% of the time. In contrast, their white counterparts are seldom incorrectly flagged.

Turnitin, a platform widely used across the nation to scan assignments for AI, has a 61.3% false positive rate for non-native English speakers. When learning a new language, simple and predictable language structure is best. That same format is what AI checkers look for.

Neurodivergent students are also prone to receive false positive ratings across all platforms—this manifests in two ways. Online proctoring tracks eye and body movement, reporting behavior that diverges from the norm as positive for cheating. Neurodivergent individuals typically move differently than neurotypical people, and thus, are flagged far more often. Secondly, these students often have a reliance on repeated phrases, terms, and words. This reliance is called “compositional masking” where neurodivergent individuals learn pattern recognition rather than prose. It is these patterns that get flagged

These injustices are not an accident! They expose a systematic failure in the creation and implementation of platforms that utilize Artificial Intelligence (AI).

Artificial intelligence (AI) has spread into every facet of our lives. Most notably, the education sector has adopted AI models for various academic systems, including grading, plagiarism detection, and student assessment. The rapid implementation of these systems has made it difficult for governance to catch up and insist on some level of oversight, with fewer than 10% of schools that employ AI worldwide having formal guidance on generative AI. When integrated without structural oversight, AI models are likely to perpetuate the stereotypes embedded in their programming. Research conducted by Students for AI Transparency (SAIT) reveals how current technical fairness metrics fail to capture deeper forms of oppression embedded in algorithmic systems, with non-native English speakers, neurodivergent users, and students of color being disproportionately impacted compared to their counterparts.

In efforts to break this cycle, SAIT is working to pass Illinois state legislation for public institutions, from K-12 to public universities. In addition to pushing reform on the state level, we are building coalitions at private universities to make this movement nationwide. By signing this petition you support our initiative that will require schools to: 

1. Be transparent about present and future AI models they utilize, and the ways that they impact the students by making this information accessible to students and parents.
2. Conduct independent auditing where third party companies test these models to scan if and where bias is embedded.
3. Reform platforms that have been flagged for bias accordingly.
4. Implement committees in each district who have oversight and say in existing and future partnerships with AI companies to ensure that these initiatives are followed.
5. Face consequences if aspects of these terms are violated or manipulated to continue the usage of harmful platforms.

By signing this petition you stand with SAIT in saying NO to harmful bias that has already inflicted harm in the lives of countless students.