Using AI to generate Research Publications

[SSerenity]

Extremely helpful article. I've been doing this during Residency for a while now, mostly for case reports but also for two reviews so far.
Just know, its not as simple as just saying "hey write me a review on sarcoidosis". It still takes human work, but once you get the work flow optimized, the effort-barrier for grinding out this type of work plummets. that is the biggest benefit IMO.

Hope this helps!

---

Members don't see this ad.

 

[deleted1168269]

I've had chatgpt write a few emails to faculty and attendings. I have to ask it to change things so often I might as well write it myself.

 

[Spikebd]

Try using Droracle.ai

 

[SurfingDoctor]

You still need your own personal touches, but I’ve had it help with all sorts of fluff.

Personal statements, letters of rec, etc. It essentially makes everyone less interesting and unique… which is hilariously true. Most people, including doctors, are incredibly…

 

[futureapppsy2]

Given that AI massively hallucinates citations, a publishable research article is the last thing I'd use it for, tbh.

 

[Cornfed101]

Given that AI massively hallucinates citations, a publishable research article is the last thing I'd use it for, tbh.

There are plugins to fix that (scholar.ai), but yes make sure it is pulling from real sources

 

[futureapppsy2]

There are plugins to fix that (scholar.ai), but yes make sure it is pulling from real sources

Also, one would need to make sure the content matches up with what they are being cited for. I'd still be really wary of using AI for a review, article, ngl--so many places it could go wrong.

 

[SurfingDoctor]

Introducing Microsoft 365 Copilot – your copilot for work - The Official Microsoft Blog

Humans are hard-wired to dream, to create, to innovate. Each of us seeks to do work that gives us purpose — to write a great novel, to make a discovery, to build strong communities, to care for the sick. The urge to connect to the core of our work lives in all of us. But...

blogs.microsoft.com

Since review articles are mostly garbage anyway, I fully support any technology that reinforces their stupidity.

 

[ChordaEpiphany]

Extremely helpful article. I've been doing this during Residency for a while now, mostly for case reports but also for two reviews so far.
Just know, its not as simple as just saying "hey write me a review on sarcoidosis". It still takes human work, but once you get the work flow optimized, the effort-barrier for grinding out this type of work plummets. that is the biggest benefit IMO.

Hope this helps!

AI is absolutely fantastic for editing research writing and recycling content. It's clear, concise, and accurate. It doesn't suffer from the same issues as casual writing or creative writing, because you don't need personality in the writing, just clarity and flow. A few awesome use cases:

1) Rewriting methods from an old paper. So you did the exact same thing twice, but the publisher won't let you copy word-for-word because that's plagiarism. Throw your methods into the AI with instructions to rewrite without changing any concrete details. Then edit to ensure the methods are still accurate.

2) Editing to fit new journal word limits. Journal A rejected your 3500 word manuscript. Journal B has a 3000 word limit. Instead of painstakingly cropping out sentences, send several paragraphs through your AI with instructions to improve conciseness.

3) Writing grant sections based on published papers. Lots of grants are just research papers reframed. I've copy-pasted entire sections of my results into GPT4 and sent them back to my PI as a section of his grant. What would have been 2 hours of rewriting the same thing in new words became 30 minutes of editing.

4) Case reports. Not sure about HIPAA issues with this, but I imagine it would do an amazing job writing the technical section of a case report based on clinical notes.

Introducing Microsoft 365 Copilot – your copilot for work - The Official Microsoft Blog

Humans are hard-wired to dream, to create, to innovate. Each of us seeks to do work that gives us purpose — to write a great novel, to make a discovery, to build strong communities, to care for the sick. The urge to connect to the core of our work lives in all of us. But...

blogs.microsoft.com

Since review articles are mostly garbage anyway, I fully support any technology that reinforces their stupidity.

Review articles are basically a victory lap. A lot of them are garbage, especially when it's the 15th review on that subject published this year, but they can be immensely helpful when starting in a new field. Another great side effect is allowing academics in low-output fields to continue to focus on quality over quantity while still impressing the study section with your H-index. I know several academics whose output is basically 2-3 great papers per year but 6-8 additional reviews or perspectives. Personally I respect that approach more than pumping out 8-10 garbage studies per year. I don't think any one researcher can realistically write more than 2-3 quality papers in a year in most fields.

 

[futureapppsy2]

AI is absolutely fantastic for editing research writing and recycling content. It's clear, concise, and accurate. It doesn't suffer from the same issues as casual writing or creative writing, because you don't need personality in the writing, just clarity and flow. A few awesome use cases:

1) Rewriting methods from an old paper. So you did the exact same thing twice, but the publisher won't let you copy word-for-word because that's plagiarism. Throw your methods into the AI with instructions to rewrite without changing any concrete details. Then edit to ensure the methods are still accurate.

2) Editing to fit new journal word limits. Journal A rejected your 3500 word manuscript. Journal B has a 3000 word limit. Instead of painstakingly cropping out sentences, send several paragraphs through your AI with instructions to improve conciseness.

3) Writing grant sections based on published papers. Lots of grants are just research papers reframed. I've copy-pasted entire sections of my results into GPT4 and sent them back to my PI as a section of his grant. What would have been 2 hours of rewriting the same thing in new words became 30 minutes of editing.

4) Case reports. Not sure about HIPAA issues with this, but I imagine it would do an amazing job writing the technical section of a case report based on clinical notes.

Review articles are basically a victory lap. A lot of them are garbage, especially when it's the 15th review on that subject published this year, but they can be immensely helpful when starting in a new field. Another great side effect is allowing academics in low-output fields to continue to focus on quality over quantity while still impressing the study section with your H-index. I know several academics whose output is basically 2-3 great papers per year but 6-8 additional reviews or perspectives. Personally I respect that approach more than pumping out 8-10 garbage studies per year. I don't think any one researcher can realistically write more than 2-3 quality papers in a year in most fields.

I definitely agree that AI has places where it could be really helpful, and those are some great examples I would agree with! I just don't think having it do your lit review for you would necessarily be one of them given the issues with hallucinations, citation accuracy, etc. (Also, when I was thinking review papers, I was thinking of systematic reviews, because we don't have many non-systematic reviews in my field, and when we do, they tend to be in book chapters or a small handful of specific journals that publish clinically-focused general reviews. I agree that they can be a great jumping off point for a new topic if nothing else!)

 

[SurfingDoctor]

Review articles are basically a victory lap. A lot of them are garbage, especially when it's the 15th review on that subject published this year, but they can be immensely helpful when starting in a new field.

Except the AI can read the entire field, provide global or detailed content and be asked pointed questions for which it provides concise answers.

Of course, this is currently dependent on good prompts, but AI is going to replace review articles because frankly, it’s better at culling data than humans.

I went to an AI conference the other day and the whole premise is that AI can replace mundane and tedious tasks to allow people to be more innovative. Review articles fit that bill perfectly.