Psychology

• Never use AI as your only resource, even for a small query - you always want the articles best for your purpose even if you only need a couple
• Use an AI tool in conjunction with the authoritative Databases in the field
• Use more than one AI tool -Compare results
• Know what the AI tool’s Knowledge Base is - Where & How is it getting its info
• Learn effective prompting techniques - 
  • Give it Context
  • Question it
  • Ask it to Self Evaluate
• Verify Results -
  • Follow to the actual article - even if you are going to use the summary - to make sure it is real
  • Check claims in multiple sources / tools
  • My experience has shown me that the less I know about the topic, the more accurate the results seem
• Know that an AI tool isn’t always faster -Just because it is new and shiny, sometimes traditional tools are faster as well as more accurate.

In my experience, AI Tools do not work very well for Literature Reviews on their own.

However, they can be a good addition to using discipline specific academic databases. 

In most fields there are discipline specific database includes considerably more of the scholarly literature in that field.  This means that even if you just need a few Articles, you have a better chance of finding GOOD articles, related to your specific topic, using a tool specifically designed to search the scholarly literature.  

This is for 2 reasons:  

1. Knowledge Base

• AI Tools have a limited Knowledge Base - Each AI tool has a knowledge base that it has been trained on and that it uses to answer your questions. 
  • AI tools do not have access to all, or often even a significant portion, of the scholarly literature in a field, 
  • Free versions of tools sometimes have even more limited access than paid versions
• Scholarly Databases have extensive Knowledge bases - Each scholarly database has created its universe of the scholarly literature and tells you what is included. 
  • Scholarly Databases in specific fields often include a significant percentage of the literature in a field. 
  • They do not search the public internet - they have intentionally created a knowledge base over time that includes much more than what is on the public internet  

2. Value Added Subject Indexing -

• AI Tools do not add subject indexing to the citations that they have access to. They can recombine things to generate new text, but for searching, they can only use what is in their knowledge base.  In most cases this is limited to Titles and Abstracts.  There are no Subject headings, or added keyword lists. 
• Scholarly Databases do add subject indexing to their citations -they add discipline specific subject headings and tags for things like Document types, Substances used, or Population Groups. 
  • This provides many more access points for any given article - so it is more likely to come up in your search.
  •   Example - 
    • A search in an AI tool or Google for articles using the keywords -  Health Equity Race - would  only find articles that used those terms in the title or abstract.  Many relevant articles would use different terminology in their title.
    • A search in a discipline specific database like PubMed - gives you many more relevant articles because they have a subject heading for Health Equity - that is applied to articles on that topic regardless of what words were in the title.  

Use a recommended Academic Database for research anytime you need:

• All the articles on a topic 
• The best articles on a topic 
• Narrowly focused articles
• The latest articles on a topic 
• Specific types of articles - Review articles, Synthesis articles, Patents, Clinical trials... 
• All the articles out of a specific Research Group

Does that mean never use an AI tool for Literature Searching?   No

AI Tools are getting better knowledge bases all the time, and the thing that they have that is different from the Academic Databases is a different sort of algorithm and better natural language processing.   

This means two good things:

• AI Tools often provide context and categorization to your search results - not just a list. 
• AI Tools can cause things to bubble up in your AI results that did not show up in your academic database search results.

So, consider using them in addition to discipline specific Academic Databases

Links to Resources on Effective Prompt Engineering