Over the last few years, numerous types of AI (Artificial Intelligence) tools have been developed and some are specifically designed to help you find sources. These tools are rapidly increasing in number and constantly changing and adapting as AI technology improves. This makes it impossible to keep up with what's out there! Nevertheless, below you'll get a broad overview of these tools and you'll also learn about some of their limitations.
To get an idea what sort of AI research tools are currently available you can ask Google for a list (see image below), then use the dropdown icon next to each tool to learn more about what it does. This list is AI generated and is unlikely to be comprehensive; it may also contain tools you wouldn't consider to be research tools, such as DALL-E, Character.ai, etc.
Or, you can consult human-curated lists, such as those developed by librarians at academic libraries:
How can you use AI tools to help you with your research? Explore the options outlined below!
You can ask chatbot tools such as ChatGPT, Gemini, CoPilot, etc. to help you brainstorm research topics or help you develop research questions or hypotheses.
In Lesson 6 you learned three steps for conducting effective searches:
Select the fewest, most essential keywords that represent your topic.
Then, gather synonyms for those keywords.
Finally, set up a Boolean search to incorporate all of your keywords and synonyms.
That lesson pointed out that you can ask chatbot tools such as ChatGPT, Gemini, CoPilot, etc. to help you with step 3, identifying synonyms. For example, suppose you wanted to find synonyms for biocrusts. Here is a prompt you might give to some chatbots to see what they come up with:
Please provide a thorough list of terms people have used over the years for "biocrusts" including different types of biocrusts such as algal crusts, etc.
And below you can see what those bots came up with!
You can ask AI chatbots such as ChatGPT, Gemini, CoPilot, etc. to help you build a Boolean search. There is of course, no guarantee the chatbot will do a good job of this, in fact, it might botch it so badly that the search will not work at all. However, these bots will probably get better at this task over time. Plus, if you are inexperienced building Boolean searches then the bot may do a better job than you could! At the very least it may give you ideas for how to build your search. For example, here is a prompt given to ChatGPT:
Please put together a Boolean search on biocrusts and soil moisture.
Here is what ChatGPT came up with:
"biological soil crusts" OR "biocrusts" OR "soil crusts" OR "biotic soil crusts" OR "microbial crusts" OR "cyanobacterial crusts" OR "lichen crusts" OR "moss crusts"
AND
"soil moisture" OR "moisture retention" OR "soil water" OR "water retention" OR "water availability" OR "soil hydration" OR "water content"
Not horrible, but also not great. A much better search would capture more variations of the word "biocrust" in both singular and plural forms and would look for more variations on the soil moisture theme by not restricting this concept to a phrase. Like this:
“soil crusts” OR biocrusts OR “microbial crusts” OR “cryptobiotic crusts” OR “cryptogamic crusts” OR “lichen crusts” OR “moss crusts” OR “cyanobacterial crusts” OR “cyanobacteria crusts” OR “bacterial crusts” OR “fungal crusts” OR “desert crusts” OR “algal crusts” OR “biological crusts” OR “phototrophic crusts” OR “microphytic crusts” OR “hydraulic crusts” OR “microbiotic crusts” OR “bryophyte crusts” OR “soil crust” OR biocrust OR “microbial crust” OR “cryptobiotic crust” OR “cryptogamic crust” OR “lichen crust” OR “moss crust” OR “cyanobacterial crust” OR “cyanobacteria crust” OR “bacterial crust” OR “fungal crust” OR “desert crust” OR “algal crust” OR “biological crust” OR “phototrophic crust” OR “microphytic crust” OR “hydraulic crust” OR “microbiotic crust” OR “bryophyte crust”
AND
soil
AND
moisture OR water OR hydration OR hydraulic OR hydro*
Google Scholar is the best known AI tool for finding sources on a topic, however nowadays there are many more such tools, such as Consensus, Dimensions, Elicit, OpenRead, Research Rabbit, ScholarAU, Scholarcy, Scispace, Scite, and Semantic Scholar. Here are some pros and cons of using these tools:
One advantage of many AI tools (such as those listed above) is that they are free to use, however some require a fee if you want to use more advanced features.
Many of these tools can do citation chaining and related-articles searching. (See Lesson 6 for more information about these techniques.) However, in general Google Scholar seems to do a better job at these tasks.
If you are not interested in conducting a comprehensive search, but you just want to find a few relevant sources, many of these tools will suffice.
Most of these tools (except Dimensions) are subject to the notorious "black box" problem. What that means is that you don't really have any idea how your search is being processed. Thus, you'll never have any way of assessing whether the tool did a good job of finding relevant, comprehensive results. For example, many of these tools utilize natural language searching rather than keyword/Boolean searching. That means you enter a topic as a sentence and the database then tries to interpret your sentence, and retrieve sources based on its interpretation. So, you have no idea how well it interpreted your topic sentence, including which words it thought were most important, whether it looked for synonyms for those words, etc.
Most of these tools (except for Dimensions, SciSpace, and possibly one or two others) are subject to the "Matthew effect." That means the algorithms these tools use for retrieving sources are designed such that articles that are highly cited, or clicked on a lot, tend to be delivered further up in your results. That means they will be seen more, used more, and then weighted ever more highly, in an ever repeating cycle. Meanwhile, there could be really good articles out there that for some reason have not gotten a lot of cites and/or clicks, and for this reason they will get buried further and further in your results. The saying "The rich get richer and the poor get poorer" perfectly captures the problem with the Matthew effect.
For most of these tools, searches are not reproduceable by others. Many of these tools don't even provide a total number of results (except Dimensions) so you'd never know if you were getting the same results each time you conducted the same search. Also, the black box issue and the Matthew effect prevent searches from being reproduceable. This is a big problem if you are conducting a systematic review (see Lesson 3) where your search needs to be reproduceable by others.
Most of these tools have few or no features for searching in particular fields of a record, or refining results, sorting results, etc.
Many of the AI tools listed in the section above have features to help you save time; for example they will extract, summarize, and synthesize content from sources so you don't have to read sources in their entirety. Furthermore, for some you can pose a question and the tool will provide an answer that it has synthesized from various sources. How well these tools summarize content and answer questions is anyone's guess, so be careful. You may want to double check to make sure the AI tool got things right, which means the tool may not save you much time after all!