Skip to Main Content

AI Research & Literature Searching

A brief overview of some AI tools available to Mac students, staff, and faculty, as well as some information on when to use them and when not to use them.

Campus Tools

As a Macalester student, you have access to several Google and Microsoft AI products.

We recommend using these tools, rather than others like ChatGPT, as our educational license builds in some extra privacy protections for users.

  • Gemini: a generative artificial intelligence chatbot developed by Google AI. Based on the large language model (LLM) of the same name.
  • CoPilot: a generative artificial intelligence chatbot developed by Microsoft. Based on Microsoft’s Prometheus model, which is based on OpenAI’s GPT-4 series of large language models (LLMs).
  • NotebookLM: a research and note-taking online tool developed by Google Labs that uses Google Gemini to assist users in interacting with their documents. Users upload documents and the LLM generates content based on these documents.

Please reach out to a librarian or ITS if you have questions. There are several privacy issues to consider before initiating an AI tool or plug in for a software you already use.


See also: AI at Macalester College

Library Databases and AI

The library’s subscription databases sometimes incorporate LLMs into their platforms. You may see auto-generated summaries of search results. 

These databases are updated on a continuous basis. This list may not be comprehensive, but is maintained on a regular basis.

  • Oxford Academic: a retrieval-augmented generation (RAG) tool that uses a ChatGPT 4o-mini large language model (LLM). Generates synopses of results.
  • ScienceDirect Topic Pages: pages created by ScienceDirect using heuristic and machine-learning. Information is compiled on a topic-by-topic basis.

When using these models, some of the results may be trained on content that Macalester may not be subscribed to. If you encounter a paywall, please submit an ILL request.

Other AI Research Tools

There are numerous other tools for using AI in research, with many more popping up every day. 

The following tools are based largely on scholarly sources, some with augmentation from models like ChatGPT in several cases.

Literature Mapping

  • Open Knowledge Maps: queries open access literature databases and uses natural language processing (NLP) methods to aggregate and display publications according to similarity in topics.
  • Connected Papers: identifies similar papers using natural language processing (NLP) methods to generates a network visualization.

Search Summaries

  • Semantic Scholar: a search engine that uses natural language processing (NLP) to generate summaries of scholarly papers.
  • scite.ai: a search engine that includes “Smart Citations,” which use natural language processing (NLP) methods to predict how a publication has been cited (i.e. supporting or contrasting evidence for the cited claim).

Other Tools 

Note: each of these tools requires an account

  • Elicit: generates large language model (LLM) summaries for research questions and retrieves papers.
  • Explainpaper: highlight text in a paper to view a large language model (LLM) generated explanation.
  • Litmaps: a retrieval-augmented generation (RAG) tool that generates literature maps based on citation networks.  
  • Rayyan: a tool for conducting systematic reviews.
  • Research Rabbit: uses natural language processing (NLP) and citation mapping to identify relevant papers.

When using these models, some of the results may be trained on content that Macalester is not subscribed to. If you encounter a paywall, please submit an ILL request.

Please reach out to a librarian if you have questions.


See also: Generative AI Product Tracker - Ithaka S+R

How do I know if an AI-generated citation is real?

Large Language Models (LLMs) sometimes provide citations for papers that don't exist, so you should ALWAYS confirm whether a source is real or not if you use AI to find literature. 

Here's a few strategies to confirm whether a source is real:

  • Search Crossref: searching the DOI in Crossref is a quick and easy way to see whether a paper exists. The Digital Object Identifier (DOI) is a persistent, standardized identifier, so it can be a helpful reality check and retrieves a reliable URL. 
  • Search Google: you can also search Google for the paper using the title and author(s). This might result in finding some papers that are not exactly the same as what's cited, but similar.

If you find the paper but encounter a paywall using either of these methods, try finding it in the Macalester library catalog or submit an ILL request

Comparing Tools

This table is meant to provide a quick overview of the differences and similarities between the tools listed above. For more information, please click on the links or reach out to a librarian.

       
Tool AI Functionality Model and/or Analysis Method Training Data
Open Knowledge Maps Topic clustering and visualization Natural Language Processing PubMed and BASE article metadata
Connected Papers Topic clustering and visualization Natural Language Processing Semantic Scholar citation data
Semantic Scholar Paper summarization and relevance rankings Machine Learning, GPT-3.5 GPT3.5-turbo-16k (OpenAI) and Semantic Scholar titles and abstracts
scite.ai Citation contextualization Deep Learning scite index (open access papers and some journals) metadata and full text
NotebookLM Chatbot, document summarization, audio and text content generation Gemini Gemini and document uploads
Gemini Chatbot Gemini Web documents, books, open code, internal Google data sources
CoPilot Chatbot GPT-4 GPT-4, Microsoft content