I think I’ve not been giving myself enough credit. I’ve observed it’s not a me-thing; it’s a human thing. We give greater honour to external factors than we do to ourselves.

 I can confidently beat my chest for how my writing compared with AI-generated writing on Grammarly scores.

How?

I drafted a piece of write-up a theory of communication I have observed, all on my Smartphone while on transit. I still love the old fashioned way of editing and proofreading with my laptop. I detect error better than when using phone. However, I fed the text into ChatGPT with a simple instruction to “Refine this write-up.” Once I saw the output, I knew refine was not the right term. Or, the reason I prefer natural writing to artificial: AI generated texts are robotic and of course, “unnatural”.

I read the output, and read my writing. I compared the two. My writing was filled with misspellings and typos. But the AI output, it didn’t feel like my writing any more. With better structuring and vocabulary, AI output lost touch with my message. It lacked the feeling that flowed in the natural writing.

ChatGPT reduced word count from 1185 to 587.

That’s when I asked myself:

“Is it that my bias against AI generated writing blinded me to judge the output negatively even when I could be better than my writing?”

“Or that my writing has some magic in it that is lacking in that of the AI didn’t have?”

I then needed a neutral, third party judge.

Grammarly.

I fed the two write-ups to Grammarly.

The reports came out.

Overall:

. My writing, a.k.a, Natural Intelligence scored: 82%.

. ChatGPT, a.k.a, Artificial Intelligence scored: 87%.

These scores are measured over all texts checked by Grammarly.  By that, AI scored higher than NI, by 5%. But Grammarly scores are more detailed. NI writing had no critical issue. AI writing had one.

That prompted me to look at it as Natural Intelligence versus Artificial Intelligence, a comparative analysis based on Grammarly reports. So going forward, I evaluate the differences in writing quality, vocabulary diversity, and overall effectiveness, as reflected by the Grammarly reports for each piece.

General Metrics:

• My Writing (Natural Intelligence): 1,185 words, 84 sentences

• ChatGPT Writing (Artificial Intelligence): 584 words, 36 sentences

Like I said earlier, the piece I wrote is longer, with 1,185 words spread across 84 sentences. In contrast, the AI-generated text is shorter, with 584 words in 36 sentences. That is 600 words difference. This difference suggests that, as a human writer, I tend to elaborate more on points. AI, on the other hand, opts for a straight-to-the-point approach. This is likely due to its design to optimize for clarity and brevity. Or, its brain (read as Intelligence) is more strictly structured, especially with the simple instruction to “Refine”; unlike my brain that can ramble around and buttress as I wish. (As I write this, my brain is already wandering, trying to debate whether I should use “Natural Intelligence” or “Human Intelligence” or “Original Intelligence”). That is by the way.

Score and Writing Issues:

• My Writing: Score 82, 50 issues (0 critical, 21 advanced)

• AI Writing: Score 87, 19 issues (1 critical, 18 advanced)

Grammarly scored my writing 82%. This means that the hasty writing I did is better than 83% of all texts checked by Grammarly. However, it flagged 48 writing issues. I had a clean slate on critical issues category. These issues likely arise from my personal style. I like to use complex sentences, phrases, and a varied vocabulary that challenge conventional grammar rules. I love to think myself a great poet, and I fuse this in my writings. Also, I use Grammarly free version. There is limited number of error I can correct with that.

On the other hand, yes, the AI-generated text scored higher at 87%. Well, only 5% ahead of mine. It had only 21 writing issues, with ONE critical issue. Remember, mine had NO critical issue.  Does the lower number of issues suggest that AI writing aligns more closely with standardized grammatical rules, probably because it is programmed to prioritize grammatical correctness? My answer is YES. “Refine” to AI means strict observation of the grammatical rules. Grammarly is practically there to check grammar.  So, it is understandable for AI to score higher than NI in the aspect.

 Vocabulary Diversity

• My Writing: 32% unique words, 33% rare words

• AI Writing: 47% unique words, 31% rare words

Well, AI defeated me in this aspect with whooping 15% in the use of unique words; but I leveled the score in the use of rare words.

My text uses 32% unique words and 33% rare words which we can call it a rich and varied vocabulary. The AI text, while shorter, achieved 47% unique words and 31% rare words.  This higher percentage of unique words shows that AI can effectively use varied vocabulary within a shorter text. I expected more of myself here, especially as human intelligence is known for creativity and personal expression. However, this may also contribute to the higher number of writing issues due to the complexity of language I used. For ChatGPT, lower percentage of rare words compared to my text suggests means that AI tends to use more common, easily recognizable words to maintain clarity and avoid grammatical errors. And remember, my writing is about 50% more than that of AI.

However, both writing performed poorly in vocabulary diversity. My score of 33% is capital “F” in Nigeria educational system. Also, 47 is a “D”. Both AI and I need to build and use more diverse vocabulary for our writings.

  Word and Sentence Length:

• My Writing: 4.8 characters per word, 14.1 words per sentence

• AI Writing: 5.5 characters per word, 16.2 words per sentence

Which should rate higher than the other: the ability to use longer words and sentences or the ability to use shorter words and sentences?

One factor that easily sells out AI-generated text is verbosity.  Grammarly reported that the average word length in my writing is 4.8 characters, and the average sentence length is 14.1 words. As obvious, I tend to use simpler, more concise words. Would you consider a sentence of 14 words to be longer, more complex sentences?  I’m aware I often tend to create nuanced and layered sentences that may sometimes sacrifice grammatical precision for rhetorical effect. My style is that if it sounds appealing and perfect like the preceding sentence but readers need dictionaries to grasp the message, I am already planting noise on the fidelity path of my communication.

Compared to my writing, the AI text has a slightly longer average word length of 5.5 characters. It has an average sentence length of 16.2 words. If you observe AI generated texts, you’d see that it favors longer words and longer sentences.  I believe that this contributed to its higher Grammarly score, as it balances sentence complexity with grammatical correctness.

For a human possessed by some inspirations, like I was while writing the texts, the flow of ideas is far more important than grammar and choices of sentences; these feature in second draft. Notwithstanding, it is still a credit to AI and machine learning.

#AI #ArtificialIntelligence #NI #NaturalIntelligence #GenerativeAI #ChatGPT #Writing #creativity #Grammarly Grammarly ChatGPT Hamict