Generative AI for Research

Generative artificial intelligence (AI) is a part of the new educational landscape. It may feel like the current conversation focuses on all the potential dangers, but using appropriate generative AI tools can help experienced scholars save time and effort with some of the more tedious parts of the research process. Here are some important considerations to keep in mind for successful usage.

Prompt engineering is the process of creating prompts, mostly in the form of written commands, that elicit specific responses from generative AI tools. Learning effective prompt engineering makes your interactions with generative AI go smoothly. More specific prompts elicit more nuanced and critical responses. One way to do this is by following Leo S. Lo’s CLEAR framework in which researchers consider the language, order and clarity of their prompts while continuously adjusting and reflecting for better outputs. Whether the tool is a chatbot or a search engine, understanding how to structure your interaction will provide a more sophisticated and successful result.

Understanding the tools limitations keeps expectations in check. Most generative AI search tools focus on scientific literature for their data, meaning their outputs for social science and humanities questions can be pretty limited. Search tools may also rely heavily on open access content, which can skew results and leave proprietary information unaddressed. Additionally, there is the regular threat of hallucinations and biases. Hallucinations are factual inaccuracies in the output from generative AI tools, while generative AI biases are similar to the racial and gender biases noted for algorithms and other types of AI. The best way to gauge the limitations of a tool is to practice using it and critically analyze the outputs for hallucinations, biases and limited information.


SMU Libraries recently hosted a workshop on six free generative AI tools that assist researchers with idea generation, searching and literature reviews (ChatGPT, ChatPDF, Consensus, Semantic Scholar, Research Rabbit and Connected Papers). The online guide contains resources on Generative AI and Research for advanced researchers. Two things to note are the chart that explains the best uses for each of the six tools and the page on prompt engineering that provides best practices and example prompts.


The best way to become familiar with generative AI tools is to try them out yourself, even if only experimenting with free versions. By recognizing prompt engineering frameworks and potential limitations, you may find the process less intimidating and more successful. Reach out to your subject librarian for more ways to get support with using generative AI.


This post was written by Julia Anderson, social sciences research librarian at Fondren Library. Julia leads our workshops on generative AI tools for research.