Ever fallen behind when taking an online course? Or, even worse, have you clicked through the...
Students outsmarting AI detectors? Here's why we need a new approach
In the ever-evolving landscape of education, technology is reshaping how students learn and are assessed. One of the most recent advancements, Artificial Intelligence (AI), brings both opportunities and challenges. As students become increasingly tech-savvy, they find ways to bypass AI detection systems designed to maintain academic integrity. This raises a crucial question: Are AI detectors the best solution, or do we need to rethink our approach to assessments in education?
Ways students bypass AI detection
Students are remarkably resourceful when it comes to finding ways around obstacles. Here are some methods they use to outsmart AI detectors.
Using AI humanisers
Tools like HIX in "balanced mode" effectively reduce AI detection and plagiarism levels by making AI-generated text appear more human. This highlights the sophistication of some AI humanisers in bypassing detection systems.
Altering punctuation and whitespace
Even minor changes in punctuation and whitespace can trick popular AI detectors. By adding extra spaces and varying punctuation marks, students can disrupt the text pattern enough to fly under the radar. This simple yet clever tactic underscores the fragility of current AI detection capabilities.
Rephrasing and rewording text
Rephrasing tools such as Spinbot, WordfixerBot, QuillBot, and Grammarly help students make AI-generated text appear more human. By altering sentence structures and word choices, these tools can effectively evade detection.
Adjusting AI parameters
Adjusting the "Temperature" parameter in OpenAI's API can produce more creative and random outputs, making detection more challenging. This technique leverages the flexibility of AI models to generate unique responses that are harder to flag as AI-generated.
Fine-tuning AI models
Some students fine-tune AI models like GPT-4 for specific tasks, such as essay writing in particular styles. This customised approach can create outputs that are more difficult for detectors to identify as AI-generated, demonstrating how students stay ahead of the technology meant to police their work.
The complex nature of plagiarism
Check out this video by Bianca Raby, our Founder and CEO, discussing students, AI, and plagiarism in the education system, which encompasses the information in this section:
What is plagiarism?
Plagiarism, at its core, is the act of using someone else's work or ideas without proper attribution, presenting them as one's own. It undermines honesty, trust, and respect for intellectual property. Completely copying someone else's work is clear-cut plagiarism and is not acceptable under any circumstances. However, the definition of plagiarism is not always straightforward.
The complications of defining plagiarism
Plagiarism can take many forms, from direct copying of text to subtle acts like paraphrasing without acknowledgment or using someone else's ideas without credit. The boundaries between acceptable paraphrasing and plagiarism can be blurry, making it difficult to navigate.
The myth of original ideas
In today’s information-rich world, the concept of truly original ideas is becoming increasingly mythical. Most new ideas are often recombinations or evolutions of existing ones. This is not to say innovation is dead—far from it. Innovation now often means making connections between existing ideas in new and unique ways. Consider the multitude of songs using the same four chords or the recurring themes in literature. Our learning, reading, and experiences continually shape our thoughts and ideas. Thus, expecting students to produce entirely original ideas without influence is unrealistic and ignores the collaborative nature of knowledge creation.
Plagiarism in the AI world
In the context of AI, the concept of plagiarism becomes even more complicated. AI models like ChatGPT generate text by drawing on vast amounts of data from the internet. This raises questions about authorship and ownership. When a student uses AI-generated content, who owns the information? Is it the AI, the creators of the AI, or the student who prompted the AI?
For example, if a student asks an AI to write an essay on international relations between two countries, the AI's response is based on aggregated knowledge. The student might edit and present this AI-generated content as their own. Is this plagiarism, given that the AI's response is an original combination of existing knowledge? The undefined nature of authorship in AI-generated content makes it challenging to apply traditional definitions of plagiarism.
Ownership and intellectual property
Ownership of AI-generated content is a murky area. While the creators of the AI models own the algorithms and data, the specific outputs generated for individual users are not straightforwardly owned by either the AI creators or the users. This ambiguity complicates the enforcement of academic integrity policies and the definition of plagiarism.
An argument against AI detectors
While AI detectors aim to uphold academic integrity, their effectiveness and ethical implications are debatable. Here are some reasons why relying on AI detectors might not be the best approach:
False positives and negatives
AI detectors are not infallible. They can produce false positives, accusing students of cheating when they haven't, and false negatives, failing to detect actual AI-generated content. This undermines trust in the system and can unfairly penalise honest students.
Limiting creativity and learning
Over-reliance on AI detectors can stifle creativity and hinder learning. Students might avoid using innovative tools and techniques out of fear of being flagged, missing out on valuable learning opportunities.
Ethical concerns
The use of AI detectors raises ethical concerns, particularly around privacy and surveillance. Constant monitoring can create a distrustful environment, impacting the student-teacher relationship and the overall learning experience.
Why educational institutions need to adapt assessments
To truly prepare students for the future, educational institutions must rethink their assessment strategies. Here’s why:
Embrace AI as a learning tool
Instead of viewing AI as a threat, educators must integrate it into the learning process. Teaching students how to use AI responsibly can enhance their research, critical thinking, and problem-solving skills. For example, a history teacher could assign students to use AI to gather sources and draft an essay outline, then refine their arguments and analysis on their own.
Foster critical thinking
Assessments need to focus more on critical thinking and analysis rather than rote memorisation. By designing assignments that require students to apply concepts, analyse data, and synthesise information, educators can encourage deeper learning. For instance, instead of a multiple-choice test on a book, an English teacher might ask students to write an essay comparing the book's themes with current social issues.
Promote originality and creativity
Encouraging students to bring their unique perspectives and creativity into their work can reduce the temptation to rely on AI. Project-based learning, open-ended questions, and interdisciplinary assignments can foster originality. Imagine a science project where students must create a sustainable solution for a local environmental issue, requiring them to apply scientific principles creatively to a local context
Encourage reflective practices
Incorporate reflective practices into assessments to help students think critically about their learning process. This can include journals, self-assessment forms, and peer reviews. For instance, after completing a project, students could write a reflection on what they learned, the challenges they faced, and how they overcame them. This encourages metacognition and helps students internalise their learning experiences.
Blend traditional and digital methods
Blend traditional assessment methods with digital tools to create a more comprehensive evaluation system. For example, a literature class could use traditional essays alongside digital storytelling tools. This approach caters to different learning stages and provides a more holistic view of student capabilities.
As students continue to find ways to bypass AI detection, it's clear that traditional methods of maintaining academic integrity need to evolve. Rather than relying solely on AI detectors, educational institutions should embrace a more holistic approach to assessments. Check out our Oppida LIVE, 'How AI is Transforming Education: A Live Webinar with Oppida', where this is discussed further. By fostering critical thinking, creativity, and responsible use of AI, we can prepare students for a future where technology and human ingenuity go hand in hand.