AI and Writing
- AI has many uses in academia and healthcare but it is not a good source to find reliable information and you should never trust citations generated by AI.
- Any questions you ask or information you provide will be integrated into the AI model, therefore you should never share personal/private information with AI tools.
Unreliable Data
- AI models rely solely on the data they have been trained on
- Can be biased, out of date, or completely incorrect
- Current AI models cannot access subscription resources (i.e. most peer-reviewed journals & articles)
- Most AI programs take user prompts into their data model meaning user questions might lead AI to include erroneous information taken from users
From Data quality and artificial intelligence – mitigating bias and error to protect fundamental rights, FRA
Plagiarism
- Trained on large sets of existing data, AI can take phrases, sentences, and core-concepts word-for-word directly from trained data
- Can create outputs with ideas and concepts taken from copyrighted sources with improper or no citations at all
- For authors or artists who do not wish for AI to use it, there is no way to remove their works
- Because most AI models integrate user prompts and questions into their training, your personal/private information could be included and shared with others
From AI and Plagiarism, Excelsior OWL
Hallucinations
- Generated AI responses containing false or misleading information
- Patterns, concepts, objects, etc. that are nonexistent or imperceptible to humans may be detected and cause the AI to create nonsensical or altogether wrong output
- Citation and reading lists may look extremely accurate but be entirely fabricated
- May even include actual author names and journal titles associated with the topic of interest
From What Are Hallucinations, IBM
For more info regarding common pitfalls of AI
