5
SciPoll 639: Using AI in Science Publication
In the instance that AI is used does it require human review?
Results
(238 Answers)
Answer Explanations
- Yesuser-284533Yes, AI is not mistake proof - sometimes it halucinates, sometimes its math ability is suspect, sometimes the grammar is not amazing. Always review.
- Yesuser-964888Should have a human review to check the relevance.
- Yesuser-637083Anything generated by AI should definitely be reviewed by a human with expert knowledge.
AI should be harnessed to make our lives easier and to remove tedious or time-consuming tasks, but the output should ultimately be checked and verified by a human. - Yesuser-397183I don't think AI is yet writing scientific papers or even emails that don't require substantial editing
- Yesuser-290459it's hard to know whether AI written things are true
- Yesuser-650602Yes. It is often wrong
- Yesuser-574398AI can provide interesting insight and perhaps some answers. It can also create garbage. Only a human expert would know the difference.
- Yesuser-615884Always!
- Yesuser-762237AI-generated content requires human review to ensure accuracy, quality, and integrity in the publishing process. Human review is critical for several reasons:
1. While AI systems are advanced, they are not infallible. They can misinterpret data, generate errors, or produce results that lack contextual nuance. Human review ensures that the information and analysis provided by AI is accurate and valid. Researchers can review AI-generated results, correct any errors, and ensure that the results are consistent with research goals and existing knowledge.
2. The quality of scientific publications must meet high standards, and AI tools, despite their capabilities, cannot fully guarantee this on their own. Human reviewers can assess the coherence, readability, and overall presentation of AI-generated content. They can make necessary adjustments to improve clarity, logical flow, and adherence to publication guidelines, ensuring that the final manuscript is of the highest quality.
3. AI cannot understand the broader context and interpret results with the depth and insight that human researchers possess. Human review allows for the incorporation of expert judgment, contextual knowledge, and nuanced interpretation that AI cannot provide. This is especially important in complex or interdisciplinary research, where contextual understanding is key to accurate interpretation.
4. Ensuring ethical integrity in research and publication is paramount. Human review is essential to verify that AI use adheres to ethical standards, such as avoiding biased interpretations, ensuring proper attribution, and maintaining transparency about the AI's role in the research process. Researchers can also ensure that AI-generated content does not inadvertently include sensitive or inappropriate material.
5. Researchers are responsible for the content they publish. Human review ensures that they take full responsibility for the work, including any content generated by AI. This accountability is critical to maintaining trust and credibility within the scientific community. By thoroughly reviewing AI-generated results, researchers can confidently stand behind their publications.
In summary, while AI can significantly improve the efficiency and scope of research tasks, human review is essential to ensure that AI-generated content meets rigorous standards of scientific accuracy, quality, and ethical integrity. The collaboration between AI and human expertise ensures that the final publication is both innovative and reliable. - Yesuser-138703have no doubt
- Yesuser-276088Always need human check of source data verification
- Yesuser-475346For the same reasons as mentioned above. It's good to use to generate initial text, but this needs to be reviewed for scientific accuracy since the machine cannot identify this consistently at the current time.
- Yesuser-907425Everything requires human review.
- Yesuser-740731AI is artificial, and it cannot judge or interpret anything based on true evidence
- Yesuser-81297Yes, I believe that AI supervision is important since you do not know where the AI is "competent". Please, remember that AI is just "interpolating" (and cannot extrapolate).
- Yesuser-245397Absolutely, AI has to have human input to make sure that the meaning is correct.
- Nouser-404499not as long as the use of AI is stated in the methods and the summary sentence
- Yesuser-905470AI is prone to mistakes. It is important that a human reviews the final work to prevent mistakes getting into final publication as these works as used in important fields
- Yesuser-638389Absolutely, yes as AI makes errors.
- Yesuser-52862If credit is to be given to the overall output of human + AI tool, then the human must take responsibility of the overall quality of the output. Therefore, human review is a must.
- Yesuser-840552Yes, it is essential that human review is carried out in the instance that AI is used to cross-validate the results it generates which can compromise the integrity and credibility of the scientific publication.
- Yesuser-543438AI still makes mistakes, we don't want inaccurate information being published.
- Yesuser-678105AI may use incorrect literature.
- Yesuser-598239of course
- Yesuser-14860AI is not perfect, so some errors can be present
- Yesuser-173340In my view this is absolutely necessary. I have a good example, I sent a scientific paper in a journal, the editor let the computer to look for plagiatorism. I had terrible problem to push the paper to be published, the editor had certain % of acceptable plagiarism and I was unable to meet that. Sure not, because the AI marked as plagiarism for example the addresses of all authors, in experimental part it were almost all descriptions of standard procedures used, which is very difficult to change without loosing the exactness, etc. In my view, this is unimportant embroilment but I would be nervous if AI should decide about e.g. a declaration of war or select a procedure for complicated surgery
- Yesuser-87331Always, AI usually has errors due to "hallucinations"
- Yesuser-979715Yes, definitely, as stated above.
- Yesuser-911600Human knowledge is necessary
- Yesuser-287804Human should definitely be involved in the verification of research findings and their publication.
- Yesuser-669208Absolutely Yes.
- Yesuser-684526See above responses.
- Yesuser-597118human overview is required since AI cannot be held accountable for what is written, furthermore it has been proven that AI can be innacurate.
- Yesuser-507408Any blind usage of AI is to be warranted.
- Yesuser-445218Hallucinations
- Yesuser-696023This is a MUST! AI might give a first draft of an English text for non-native English speakers, but it has to checked again and again. Even human translators have to be checked; I have seen enough translations which have completely lost the intention of an article, because translation also requires some understanding of the topic, not only of the language!
- Yesuser-596010Human supervision is critical.
- Yesuser-445202Todas as IAs que já utilizem possuem falhas
- Yesuser-554477Absolutely!!
- Yesuser-673903Assuming the researcher actually knows what was done... All AI??
- Yesuser-544555Definitely!
- Yesuser-514238Absolutely required.
- Yesuser-744008Always!
- Yesuser-67936AI should only be used to gather data, if at all, the human should review the findings and use them as any other data to generate ideas and interpretations of the data.
- Yesuser-123746Sure , need verification for falsified and improper content.
- Yesuser-689910AI should be used both to assist in writing manuscripts (for which authors are ultimately responsible); and they should be used in assisting editors in peer-review.
- Yesuser-266855Always, it depends on the quality of sources used for the content and may be biased if there is much non-scientific media on the topic.
- Yesuser-874889The human review make the paper a live
- Yesuser-799639Mistake in AI product might affect the human reputation/credibility/acceptance
- Not Sureuser-41956I can see AI helping search QA data sets for specified information. I oppose AI using data to create a de novo report or datum.
- Yesuser-480376Of course. The work is intended for human use, so humans should review it.
- Not Sureuser-935064The power of AI is that it can do things that humans are unlikely to do, at least in a reasonable time frame. For example, pattern recognition from billions of data entries requires AI. However, at least with current technology, AI output should be viewed as generating hypotheses to be pursued further, for example, to determine if there are causal relationships rather than correlations.
- Yesuser-731405Definitely!
- Yesuser-902187Authors are responsible for the content. AI is not yet reliable for use without review.
- Yesuser-779737AI makes lots of mistakes, and the purpose of the peer review process is to make sure the content is correct. So of course humans are still needed to review any content, AI created or otherwise.
- Yesuser-383159AI responses can be unreliable or shallow
- Yesuser-834001Of course this needs human review.
- Yesuser-200863Oh yes.
- Yesuser-802778It is a must
- Yesuser-642018Yes. AI requires human review. AI tools do not replace human insight and perogative. AI could sometimea misunderstand concepts so human review is essential
- Yesuser-809947Getting error is common everywhere. Its better to have a review
- Nouser-785535No if AI is used for a check on the manuscript in the finalisation step (see point 4)
- Yesuser-987162human review is a must as not all data would be correct or required by the researcher
- Yesuser-146796AI makes mistakes, and should be reviewed by humans as well
- Yesuser-548892Some of the answers generated need to be refined. So someone should review and edit the document. Not everything generated by AI should be used
- Yesuser-156666The final check shoud be done by human
- Yesuser-560974Absolutely yes, because AI may generate incorrect or highly inaccurate content.
- Yesuser-79668Should be reviewed for accuracy
- Yesuser-111454This is to find out whether the information generated by AI is true and correct.
- Yesuser-478855Ultimately, human authors are responsible for the paper, and they must review whatever the AI has conjured
- Yesuser-885754It's crucial to have a professional expert in the field filter the AI.
- Yesuser-583633AI brings a new perspective to bear in reviewing and writing about complex subject matter. However, the responses from AK are often incorrect or misinterpretation.
- Yesuser-764272The facts need checking
I actually think peer review should be done by both ai and people.... people should just be checking science...ai should be checking grammar, spelling, and plagiarism - Yesuser-887652AI output merely resembles authorship.
- Yesuser-284769Human review is mandatory.
- Yesuser-765807When AI is used in the process of writing a publication, it should undergo human review. While AI can assist with tasks like summarization, data analysis, and grammar checking, human judgment remains crucial. Human reviewers ensure that the content aligns with scientific rigor, ethical standards, and the intended message.
- user-232098To ensure accuracy, relevance, and adherence to scientific standards, as well as to verify that the AI outputs are free from biases and errors.
- Yesuser-15300Most importantly
- Yesuser-634057Yes since scientific publications are cathered to very technical crowds it is highnly advised to keep a human expert in the loop to vet that the language model will not hallucinate or misinterpret the real results
- Yesuser-952116Always, otherwise the author shouldn't be able to claim the work their own.