Is it Safe to Use ChatGPT for Writing a Statement of Purpose (SOP)?

Introduction

With the advent of advanced AI tools, students are increasingly turning to solutions like ChatGPT to assist in crafting compelling academic essays and Statements of Purpose (SOPs). These tools offer a new level of convenience and sophistication in composing structured, articulate, and personalized narratives. However, this technological aid raises an important question: Is using ChatGPT for such critical applications safe? Specifically, do universities have the means to detect and potentially reject AI-assisted applications?

This blog post delves into the safety of using ChatGPT for writing SOPs, examining the potential for detection by colleges and discussing strategies to minimize this risk. Our discussion is geared towards students seeking clarity and assurance on whether leveraging AI will jeopardize their application's authenticity and acceptance.

How ChatGPT works: Generation vs Plagiarism

ChatGPT, created by OpenAI, is an advanced large-language model that uses a form of artificial intelligence called deep-learning to generate text. This technology is trained by analyzing a vast collection of written material from a variety of sources, learning to predict the next word in a sequence in a way that’s similar to how a human might write. Its training involves processing large amounts of text data to understand language patterns, grammar, and information, which allows it to compose text that is coherent and contextually relevant.

Unlike plagiarism, which is the direct copying of someone else's work, ChatGPT synthesizes information based on patterns learned during its training. This means that the content generated by ChatGPT is original in its construction, significantly reducing the risk of being flagged by plagiarism detection software commonly used by educational institutions.

For students, this distinction is crucial. Using ChatGPT to help with your SOP can be a safe approach as long as the content remains unique and personalized to your experiences and aspirations. The model does not replicate existing texts verbatim but instead produces a tailored piece that reflects a synthesis of the input provided.

Detectability of AI-Generated text

The discernibility of text generated by AI has been a contentious topic. OpenAI itself had to discontinue an initiative aimed at distinguishing between human and AI-authored text. Citing a statement from OpenAI, an article in The Verge highlighted the organization's acknowledgment of the classifier's low accuracy and its propensity for false positives. OpenAI's efforts now seem to be pivoting towards developing tools to identify AI-generated audio or visual content, with no clear strategy for text as of yet.

GPTZero is another tool that attempts to discern between human-authored and AI-generated text, utilizing metrics like burstiness and perplexity to analyze writing patterns. However, this tool has faced scrutiny due to its inconsistency in accurately classifying texts. Doubts have been raised about the technical feasibility of reliably detecting text created by advanced AI models, as noted in a paper from the University of Maryland titled "Can AI-Generated Text be Reliably Detected?". The study revealed that tools like GPTZero are not dependable in practical applications, often yielding both false positives and false negatives. Further criticism is echoed in independent analyses, including a Washington Post article acknowledging GPTZero's susceptibility to errors. An illustrative example of this imperfection is GPTZero's erroneous classification of the U.S. Constitution as a product of AI, highlighting the limitations of current detection technology.

Moreover, techniques have been developed to circumvent the detection capabilities of GPTZero. By using sophisticated prompts or performing multiple iterations of text through different AI models, the 'fingerprint' of AI authorship can be effectively masked. Even simple instructions for the AI to emulate human writing or subsequent editing by human hands can challenge the reliability of these detection tools. Such evasive tactics highlight the complexities involved in the ongoing efforts to distinguish AI-composed text from that of a human's.

Our Recommendations

Opt for GPT-4 Over GPT-3.5: GPT-4, accessible through a subscription costing around $20 per month, significantly surpasses its predecessor in generating text that feels more humanlike, creative and less mechanical. The investment in GPT-4 can yield a text quality that better captures the nuances of personal expression necessary for a compelling SOP.

Use ChatGPT as a Drafting Tool: Approach ChatGPT as a means to overcome writer's block and to kickstart your SOP. Instead of submitting AI-generated content directly, consider the output as a preliminary draft. This serves as a foundation to build upon, saving you time and effort in creating an initial structure for your document.

Rigorous Proofreading: While ChatGPT typically avoids grammatical mistakes, it may inadvertently introduce factual inaccuracies. A meticulous review process is crucial to identify and rectify any such discrepancies, ensuring the integrity and accuracy of your SOP.

Personalize and Refine: Sentences produced by AI may occasionally seem formulaic. It's advisable to use these outputs as a conceptual guide while infusing your own unique style and creativity to enhance authenticity and individuality in your writing.

Employ Complex and Detailed Prompts: To elicit a more tailored and nuanced SOP, provide ChatGPT with comprehensive background information such as academic achievements, extracurricular activities, work experiences, motivations, and goals. Clarify why you are drawn to the particular university and what your future aspirations are. The more detailed and specific your prompts, the more personalized and effective the AI's response will be. Direct the AI to emulate a humanlike tone and advise it to avoid sounding mechanical, thereby ensuring the final product resonates with genuine personal engagement.

Conclusion

In conclusion, while it is possible to use ChatGPT safely for writing an SOP, it is essential to do so with ethical consideration and awareness of institutional policies.