How to use Elicit Research Artificial Intelligence for Better Results — Part 1

Ric Raftis
5 min readDec 10, 2022
Research is always about questions
Photo by Emily Morter on Unsplash

WARNING: This process can take you on a never-ending journey down research rabbit holes. It is unnecessary to go any deeper than one level, though, to get quality insights. Only you can decide when enough is enough.

Introduction

The video included here for reference is about twelve months old now as I write this article in Dec 2022. That said, it is very well presented and explains the use of Elicit very well. You will note, however, that the interface for the platform has changed slightly from the video, so you will need to make some allowances for differences.

The intention here is to develop a templated workflow process for using Elicit. I suggest you build a template in Word or Google Docs to follow when doing research. But first, what is Elicit? Elicit is effectively a research and search assistant to find academic papers. Your research does not have to be academic to use the platform, though. You can use the platform to write books, articles, blog posts and many other mediums to provide credibility to your output. The platform is based on the Generative Pre-trained Transformer 3 (GPT3) where you input a question and Elicit will respond.

The purpose is to:-

  • Help work out your topic
  • Search through Effective resources
  • Brainstorm research questions
  • Clarify your thinking
  • Narrow thinking to be even more specific
  • Find previous work
  • Identify other works by researchers
  • Find potential advisors
  • Identify experts and advisors in the area.

Using Elicit

On opening Elicit, the screen is very clear and simply requires you to Ask A Research Question. It also provides the option to “run Elicit over your own papers” which we will cover later.

Currently, I am researching volunteering in Australia, so I will use the research question “why do Australian organisations struggle to get volunteers” as the research topic. The results page then comes up and you notice that on the left-hand side there is information about the papers with Abstract Summary ticked. On the right-hand side of the page, you can see the Abstract Summaries related to the article.

Now you have the article title, authors, year and number of citations together with the brief abstract. At the top right, there is a switch to show only articles that have PDFs. The switch can be very useful to avoid showing articles that are behind paywalls and not accessible as PDFs. It is worth looking at the initial returns, though, in case you miss some really important works if you turn PDFs on. You should have noticed that articles with PDFs show it in the article's description. Another trap for new users is that although the right-hand column is shows “Abstract”, if a PDF is available, it includes the full text in this area. At this stage, however, you cannot annotate it, so it is probably best to go to the full PDF and save it to Zotero or the like where you can do your annotations.

Overall Review

Now it is time to review your results for relevancy. There may be enough there, or you may wish to click on “Show More” at the foot of the page to generate additional resources. Before you do this, however, a way of further refining your search even at this stage is to “Star” articles that particularly interest you. When you do this, the options at the bottom change to “Show More Like Starred” and “Clear Unstarred”. I like to clear the unstarred articles before showing more.

With the last lot of papers displaying you wish to consider, you will note a few more options you have. By clicking on the Filter button (top right of screen), you have the option to query keywords, set a date where you only want papers published since then and also the study type.

Reviewing an Individual Paper

After clicking a paper on the main search screen, it will open into a modal showing details including the abstract from the paper. In the left-hand column, if you scroll down, you can find possible critiques of the paper which can be useful for developing alternate arguments.

At the very foot of the left-hand column, you can ask about the paper. A question could be anything, but examples are “What recommendations does the paper make?”, or “Did the article offer any solutions?”. The answers will appear highlighted in the Abstract area. The platform keeps these questions and if you scroll down in the left-hand column, you will see the Custom questions and answers.

Extracting the Results

There are three ways you can extract the results of your research.
1. Copy and paste from the screen to your preferred medium;
2. Export to a CSV file;
3. Export to a BIB file and import it to Zotero. (I tried MyBib also, but it didn’t work).

If using option 1, then you can do this to Word or Google Docs. The HTML behind the site has these results structured as a table, so a copy and paste will render the results as a column. It works well for Word or Google, but trying to copy to a markdown file in Obsidian was a disaster and would require additional formatting time.

The second option will export everything to a CSV file. It is disappointing you cannot filter what you want to export, so you end up with the lot. The CSV file comes in raw, so you will need to organise your column widths to make it all readable. You can also delete the ones you don’t require.

The third option worked very well with Zotero using the Bib file export. In fact, it worked the best of all the options except for one minor issue. If the article has a PDF version, Zotero is not attaching this. Using Zotero direct and saving the article with the web extension, it adds the PDF. I have written to the developers about this issue. Perhaps the solution is to save the articles one at a time from the web as you review the content.

If you choose to go into considerable depth, the next steps follow in a separate article. However, if Elicit’s answers provide you with the amount of information you have generated, you can stop here and work with what you have found.

Conclusion

I have found Elicit to be an amazing research tool and look forward to exploring further. The power of analyzing different areas with the tasks is extremely useful. The extended areas which will I will cover in the next article are useful if you want to find additional questions to ask and find experts on the subject.

**Read Part 2 to continue the journey.

Thanks! I really appreciate you reading my work. If you enjoyed this article and found it useful, you can clap (up to 50 times) and follow me for more articles. If you spotted something that wasn’t clear or wish to add to the content, please leave a comment below.

You can also join Medium for thousands of Member Only stories through my link which supports my work.

--

--

Ric Raftis

Philomath who loves to write on Obsidian, PKM, AI, Community and community leadership https://ricraftis.au