Student gets zero points due to AI testing app error

Many universities in the US use artificial intelligence (AI) testing applications to prevent students from using AI to write essays. However, there are many cases where the testing tools are incorrect, causing students to receive zero points.

In a specific case, Moira Olmsted (24 years old), a student at Central Methodist University (USA), told the story of receiving a score of 0 after the AI ​​testing application evaluated her essay as "an AI-generated article", according to Bloomberg news agency.

The rate of AI testing applications “misunderstanding” ranges from 1% to more than 10%

The professor claimed that an AI-based test had previously flagged her writing. Olmsted disputed the findings, claiming that her writing was “engineered” because she has autism spectrum disorder, not because the writing was generated by AI. The school later overturned her score and passed the course.

In another case, Ken Sahib, an American student who grew up in Italy, said he was shocked when his paper received a zero in a computer networking course at Berkeley College in New York City (USA).

The lecturer explained: "Every tool I tried gave the same result: AI generated this essay." However, Sahib asserted that he spent his teenage years studying in Italy, so English was only a second language and his writing would not be as natural as a native speaker, and also asserted that the AI ​​test app gave an incorrect assessment. After the male student protested, the school changed the essay's grade.

About two-thirds of more than 450 teachers surveyed in the US said they regularly use AI testing apps, according to a report by the Center for Democracy and Technology (USA) in March. Educators say these tools help them determine whether AI has generated sentences, paragraphs or entire student papers.

A series of tools such as Turnitin, GPTZero and Copyleaks are being used in many high schools and universities in the US.

Tech companies claim that AI testing apps are up to 99% accurate. However, Bloomberg conducted a test with 500 essays before ChatGPT launched in late 2022. The results showed that the rate of AI testing apps “mistaking” student-written essays for “AI-generated essays” ranged from 1% to more than 10%.

With this error rate, considering the large number of papers that need to be checked each year, this greatly affects the psychology of students, even creating unnecessary suspicion and tension between lecturers and students.

Specifically, after being flagged for an "AI-generated article", female student Moira Olmsted was so sensitive that she used the video recording function of her entire computer screen while writing her essay to prove it to the school.

Many students in the US say they now spend more time writing essays because they have to carefully check sentences that are at risk of being labeled “AI writing” by the checking tool.

Quillbot app rates an article as having 43% AI-generated content

A 2023 study by Stanford University found that an AI testing app was “nearly perfect” at testing the essays of American-born eighth graders. However, the app incorrectly labeled half of the essays from students whose second language was English as “AI writing.”

AI testing apps should not be used as 'judges'

Faced with that reality, some schools in the US provide students with AI checking tools to self-check their papers before submitting them.

In addition, technology companies emphasize that schools should only view AI testing apps as support tools, not as “judges” to pass judgment on students' essays.

Some students who want to save time are even forced to use “AI humanizer” tools - which can automatically edit human writing or rewrite it to pass AI testing applications.

For some educators and students, schools and lecturers abusing AI testing applications will affect the teaching and learning process.

“AI is going to be part of the future whether we like it or not,” notes Adam Lloyd, an English professor at the University of Maryland. “It’s a mistake to see AI as something we need to eliminate from the classroom or discourage students from using.” Instead of using AI testing apps, Lloyd relies on his intuition. “I know my students’ writing abilities. If I have any doubts, I discuss them openly and don’t rush to accuse students of using AI to write their papers,” Lloyd says.

Source: https://thanhnien.vn/sinh-vien-bi-0-diem-vi-loi-cua-ung-dung-kiem-tra-ai-185241028001727599.htm