What is Summon Research Assistant?
The Summon article search tool (which is also the "Quick search for articles and more..." search box on the library's database pages) now features an AI-based Research Assistant tool. It is a beta release (meaning that it is still considered to be in development), but can provide useful summaries of search results. To use it, click on "Summon Research Assistant" at the top of the Summon search page.
If you are outside the library, you will be prompted to use a Google or Microsoft account to gain access (this is to limit the number of searches done, since generative AI is more resource-intensive than regular database searching. It does not track search terms, results, etc.)
How does it work? How is it different than regular Summon search?
Regular searching matches your search terms (e.g., Monet AND Giverny) against a large database of published articles and documents and returns links to those articles, ranked in order by relevance, dates, etc. Research Assistant lets you create a prompt with natural-language questions and requests (e.g., "How did his surroundings at Giverny affect Claude Monet's work?"). It then searches Summon's very large database of articles (NOT the entire internet) and provides a written summary, based on the top five articles it retrieves. Each point in the summary also includes a link to the specific article that supports that point. Additionally, if you don't like the the overview or articles, you also can run a pre-written search in the traditional Summon interface using the "View related results" link.
Keep in mind that this is developing technology. If you get unexpected results, you may want to modify your prompt and see if it does a better job. If it still does not do a good job, you may want to try the "View related results" link to see what kind of articles regular Summon retrieves. If you get results that you consider highly inaccurate, biased, etc., feel free to report those to the library so that we can ask the Summon support staff to look into them.
How is Research Assitant Different from Other AI Tools, Like ChatGPT, Gemini, etc.?
The main difference is that those products search the entire Internet as the source (the Large Language Model or "LLM") for their summaries, while Research Assistant is working from a more focused database of published articles and curated sources. It's still very, very large, but not as freewheeling as the content published across the whole internet. This may result in more authoritative results on topics that require them.
Research Assistant is also not "conversational" the way those popular tools are. ChatGPT, Gemini, etc. understand the context of follow-up questions after your initial query (like "why is that the case?" or "provide more details") and can give follow up answers. For Research Assistant, each question is considered a stand-alone question, so you'll need to provide enough context in each prompt to have it run a unique search.
A Lot of My Articles Are "Citation Online" and Don't Have Full Text Available For Them
The Research Assistant searches a very large database of articles, many of which may be beyond the holdings of the databases the library subscribes to. If you only see the option for "Citation Online" for an article, you can click on that for additional details and a link to "Request a copy from another library (Interlibrary Loan)."
Another option is, when you first do the search, to click on the three dashes to the left side of the search box to open additional search filters. One of them is to limit to "Available online." The Research Assistant will then limit its results and summary to sources that are available full text through the library's subscription databases (e.g., Gale, Ebsco, ProQuest, etc.) or select Open Access sources (e.g., arxiv.org). This may limit the information available for generating the summary, but should increase your ability to retrieve a full text article quickly.
Can I Use Research Assistant Summaries for School Projects, etc.?
The summaries generated by Research Assistant are meant to be quick conceptual overviews of the articles retrieved by the search terms. It may not completely replace the need to read the actual articles. Consult with your teacher, organization, etc. to see what they allow in terms of relying on AI-generated summaries for assignments and projects. Also note that rules for citing generative AI sources are also evolving.