- المشاركات
- 66
- مستوى التفاعل
- 6
- النقاط
- 8
A common mistake in using artificial intelligence to write scientific papers or master's and Ph.D. theses:
Many researchers, when they want to write about a research topic, especially the *Introduction* and *Discussion* sections, rely on AI to extract papers that support certain points, and ask it to summarize how these papers support the specific point. However, this is where AI tends to go off track and gets confused, often hallucinating strongly. This applies to all AI tools I have encountered, including paid ChatGPT, Deepseek, and Perplexity.
For example: Suppose you conducted your own study and found that one AI application is not more accurate than radiologists in reading chest X-ray images. You want to write a *Discussion* comparing your findings with previous research and demonstrate that studies suggesting AI performed better actually have methodological issues, such as limited sample sizes and selection bias.
If your prompt is:
*Find three studies where AI outperforms human radiologists in interpreting chest X-rays, and summarize each in one sentence that highlights key methodological limitations such as small sample size and selection bias.*
The AI may indeed find comparative studies but will often mix up the summaries, producing descriptions that appear to answer your request about methodological issues. However, these issues are not actually present in the papers it retrieved.
I’ve seen this mistake repeated frequently among researchers. Therefore, it’s important to note that:
1. AI is useful for providing papers on a particular topic you're researching. You may indeed find papers that you wouldn’t have discovered through traditional searches in PubMed, Google Scholar, etc.
2. AI is useful for summarizing a paper, but your request for a summary must not include clues that could lead it to twist the summary in a way that’s incorrect and biased toward what you want.
3. You must read every paper you extract or at least read the abstract. It's never appropriate to ask AI to summarize, then copy/paste! Doing this turns your writing into a mix of errors and could put you in an embarrassing position!
4. Write yourself first, even if your English is weak, but organize and structure your ideas. Afterward, AI can be useful in refining the phrasing and improving the smoothness of transitions. However, it is never acceptable to slack off and expect AI to complete the whole task for you.
Many researchers, when they want to write about a research topic, especially the *Introduction* and *Discussion* sections, rely on AI to extract papers that support certain points, and ask it to summarize how these papers support the specific point. However, this is where AI tends to go off track and gets confused, often hallucinating strongly. This applies to all AI tools I have encountered, including paid ChatGPT, Deepseek, and Perplexity.
For example: Suppose you conducted your own study and found that one AI application is not more accurate than radiologists in reading chest X-ray images. You want to write a *Discussion* comparing your findings with previous research and demonstrate that studies suggesting AI performed better actually have methodological issues, such as limited sample sizes and selection bias.
If your prompt is:
*Find three studies where AI outperforms human radiologists in interpreting chest X-rays, and summarize each in one sentence that highlights key methodological limitations such as small sample size and selection bias.*
The AI may indeed find comparative studies but will often mix up the summaries, producing descriptions that appear to answer your request about methodological issues. However, these issues are not actually present in the papers it retrieved.
I’ve seen this mistake repeated frequently among researchers. Therefore, it’s important to note that:
1. AI is useful for providing papers on a particular topic you're researching. You may indeed find papers that you wouldn’t have discovered through traditional searches in PubMed, Google Scholar, etc.
2. AI is useful for summarizing a paper, but your request for a summary must not include clues that could lead it to twist the summary in a way that’s incorrect and biased toward what you want.
3. You must read every paper you extract or at least read the abstract. It's never appropriate to ask AI to summarize, then copy/paste! Doing this turns your writing into a mix of errors and could put you in an embarrassing position!
4. Write yourself first, even if your English is weak, but organize and structure your ideas. Afterward, AI can be useful in refining the phrasing and improving the smoothness of transitions. However, it is never acceptable to slack off and expect AI to complete the whole task for you.
