AI in our academic settings 

Artificial intelligence (AI) is rapidly transforming the world around us, and education is no exception. AI tools like ChatGPT and Perplexity can be valuable resources for students, assisting with research, writing, and critical thinking. However, using these tools responsibly and knowing their limitations is essential. 

This guide will explore how to leverage AI’s capabilities while maintaining the integrity of your work. 

Understanding limitations 

AI models, including ChatGPT, generate text based on patterns learned from large datasets. However, they can produce plausible yet incorrect information, a phenomenon known as “AI hallucination.” This occurs when the AI fabricates details or generates responses that sound credible but lack factual accuracy. Users should be aware that: 

• AI outputs may not always reflect the most current information or research. 

• The information provided should be cross-verified with reliable sources to ensure accuracy. 

Della Nagle, a Spanish and English teacher at Berna Middle School in Texas, highlights a crucial point: 

“Unfortunately, students… are not able to understand that by using AI, they are cheating themselves out of learning how to write in their own words and voice.” 

Cross-referencing AI 

To mitigate the risk of relying solely on AI-generated content, students can implement a cross-referencing strategy efficiently by: 

• Allocating dedicated time slots for verification within their study schedules. 

• Using tools like reference management software to track sources quickly. 

• Prioritizing authoritative sources such as peer-reviewed journals, government websites, and established publications. 

This structured approach ensures students can verify content effectively without becoming overly time-consuming. For instance: 

• Use multiple AI tools: Compare outputs from different AI models (e.g., ChatGPT vs. Perplexity) to identify discrepancies and corroborate information. Perplexity, for instance, is designed to provide accurate and sourced answers, making it a valuable complement to generative models like ChatGPT. 

• Fact-check against reliable sources: Always verify claims made by AI tools against established databases, academic journals, or trusted websites. 

Kelly Gibson, a high school English teacher, shares her experience: 

“Instead of spending all my time trying to catch students using AI to cheat, I’ve found ways to implement it in the classroom. I think educators will increasingly learn how to adapt to AI.” 

Citing AI 

When using AI-generated content in academic work, proper citation is crucial. Here’s how to reference tools like ChatGPT according to APA and MLA styles: 

APA citation example: 

• In-text citation: (OpenAI, 2024) 

• Reference list entry: OpenAI. (2024). ChatGPT (May 24 version) [Large language model]. https://chat.openai.com/ 

MLA citation example: 

• In-text citation: “ChatGPT generated the following distinctions…” (OpenAI). 

• Works Cited entry: OpenAI. ChatGPT. May 24 version, https://chat.openai.com/. 

Critical thinking 

Students should be encouraged to approach AI-generated content critically. For example: 

• Evaluate the source: Assess the credibility of the information provided by the AI tool. If an output includes references or citations, verify their authenticity. 

• Engage with content: Rather than accepting AI responses at face value, students should analyze and question the information presented. This practice fosters more profound learning and critical thinking skills. 

For instance, when using AI to generate an essay outline, students should compare the AI-generated structure with traditional guidelines provided by their instructors or textbooks. 

 They should ask themselves: 

“Does this outline cover all necessary points? Are there any gaps in logic?” 

Additionally, students can practice critical thinking by exploring multiple perspectives on a topic suggested by AI, ensuring they do not rely solely on one source for their understanding. 

Matt Miller, an educational speaker, emphasizes the importance of creativity in learning: 

“Let’s provide creative, authentic ways for students to show what they know. Creativity in learning can help students with motivation. It can help them make use of their own unique talents and skills. It can help them feel seen and heard and noticed.” 

Using grounded models 

Using grounded models that provide citations and are less prone to hallucination can enhance reliability: 

• Tools like Perplexity not only deliver answers but also include sources for verification, making them more trustworthy for fact-checking purposes than generative models that may create fictitious references. 

• iAsk.ai is another AI tool that provides accurate and factual answers to user questions. It boasts high performance on benchmarks like MMLU and TruthfulQA. 

Key benefits of using grounded models 

• Enhanced accuracy: Grounded models are trained on massive datasets of text and code, along with factual information from credible sources. This results in a higher degree of accuracy compared to generative models. 

• Reduced bias: Grounded models are designed to minimize bias by considering information from various sources. This helps to ensure that the information you receive is objective and unbiased. 

• Improved transparency: Grounded models often provide citations and sources for their answers, allowing you to verify the information and learn more about the topic. 

Looking ahead 

While AI tools such as ChatGPT can enhance learning and productivity, they must be used carefully. Key takeaways include: 

• Understanding the limitations of AI, such as inaccuracies and biases. 

• Implementing efficient cross-referencing strategies to verify information. 

• Properly citing AI-generated content in academic work. 

• Applying critical thinking to analyze and validate AI outputs. 

By following these best practices, students can effectively integrate AI into their learning without compromising the integrity of their work. 

Can you tell what’s AI authorship? 

Key indicators for written/made by AI content 

As AI-generated content becomes increasingly prevalent, it’s crucial to be able to distinguish between human-created and AI-generated material. Here are several key indicators to help identify AI-generated content: 

Text indicators 

Unnatural language patterns 

• Overuse of common words like “the,” “it,” or “is.” 

• Repetition of specific phrases or vocabulary. 

• Unusual errors or awkward turns of phrase (e.g., referring to trees as “Arbor Day notables”). 

Lack of a personal touch 

• Absence of personal experiences or emotional depth. 

• Dry, “robotic,” or matter-of-fact tone. 

• Missing human experiences and opinions. 

Inconsistencies 

• Abrupt shifts in tone, style, or topic. 

• Has outdated or exaggerated information. 

• Fakes sources and references. 

Perfectionist red flags 

• Flawless grammar and lack of typos (human writing often contains errors) 

• Overly smooth and consistent writing style 

Visual indicators (images and videos) 

Unnatural features 

• Misplaced or missing objects, especially hands 

• Unusual smooth skin or clothing textures 

• Inconsistent lighting and shadows 

Movement and audio issues 

• Unnatural body movements or facial expressions 

• People who don’t blink in videos 

• Mismatched audio and video, including unnatural voice tones 

General tips 

• Check for watermarks indicating AI generation 

• Be wary of content that seems “too perfect to be true” 

• Verify sources and cross-reference information 

• Use reverse image searches for visual content 

Attention: AI detection tools are not always reliable, and the landscape of AI-generated content is rapidly evolving. Staying vigilant and using critical thinking skills are essential. 

Important considerations 

Identifying whether AI has generated text or content involves examining several key characteristics and behaviors that distinguish human writing from machine-generated output. Here’s a detailed breakdown of additional indicators: 

1. Unusual or over-complex sentence construction 

1. AI-generated text may exhibit sentence structures that are grammatically correct but sound awkward or overly complex. 

1. Lack of personal experience or emotion 

1. If a text lacks anecdotes, personal insights, or emotional resonance, it could indicate that an AI produced it. 

1. Overuse of certain phrases or vocabulary 

1. AI models can use specific phrases or vocabulary repetitively due to their training data. 

1. Not referencing current events (or misrepresenting them) 

1. Many AI models do not have real-time access to current information and cannot accurately comment on recent events beyond their training cutoffs. 

1. Lack of depth/analysis 

1. AI-generated content often lacks the complexity and originality found in human analysis. 

1. Inconsistent tone and style 

1. AI may produce text that shifts tone or style inconsistently throughout the piece. 

1. Errors in contextual understanding 

1. AI can misinterpret context, leading to statements that seem out of place or irrelevant within the overall narrative. 

By being aware of these indicators, readers can better discern whether a piece of writing is likely created by an AI system rather than a human author. 

Authoritative sources to be used: 

• Google Docs help center: This section provides detailed instructions on using Google Docs features, such as Version History, which helps track changes made during document editing. 

• OpenAI research papers: Offers insights into how AI language models function, including their limitations regarding context understanding and emotional depth in generated texts. 

• Originality.ai is recognized for its high accuracy in detecting AI-generated content, boasting a reported accuracy of 98%. It employs advanced algorithms specifically designed for this purpose, making it a reliable choice for educators and content creators looking to verify authorship. 

• GPTZero specializes in detecting content from popular AI models using a seven-layer detection model. While GPTZero performs well with specific AI-generated texts, it has faced challenges with mixed content. GPTZero is beneficial for educators who are concerned about academic integrity. 

• Percent Human Chrome Extension allows educators and students to easily detect AI-generated content directly from their browsers, providing a convenient option for quick assessments. 

Stay informed and always question the authenticity of content! It is up to you to make the call and think critically to understand where that information is coming from.