Generative artificial intelligence (AI) is quickly becoming a popular technology in the public sphere, as well as in higher education. There are many applications for academic research, but they come with the inevitable risks and pitfalls of emerging technologies. SJU Libraries has created this guide to help students and faculty navigate those issues and utilize AI tools to support education and research. Scroll down this page and explore the tabs on the left for advice on best practices, useful tools, prompt crafting, citation style guides, and more information and resources on the background and uses of AI.
Disclaimer: SJU does not promote or endorse any of the products or tools listed in the guide. It is meant for informational and educational use only.
AI generated image borrowed from Association of Research Libraries website article (see references).
DO:
Use your own critical thinking when considering topics and keywords, and when generating prompts, so your research and writing are original and reliable.
Use AI for brainstorming when you’re getting started with your topic, focus, and keywords.
Use AI for summarizing broad and complex topics to help understand the important concepts.
Use caution when reviewing AI responses, they may contain hallucinations, inaccurate information, and biases based on human input.
Use style guidelines for citing the use of AI in your original work.
Use the ACRL Framework for ethical evaluation, assessment, and application of AI produced content.
DON’T:
Depend on AI for doing your research or writing your paper which should be original work.
Take AI responses for granted, as they may contain hallucinations, inaccurate information, and biases based on human input.
Upload personal and/or copyrighted work, as it will be used by AI to generate future responses and may end up sharing private data and intellectual property.
Cite AI like a published article because it is not permanent, it is generated by the human user.
AI-Assisted Plagiarism: AI-generated content that resembles existing work. Since the work is not generated by the author, it requires proper citation, or at least an acknowledgment or disclosure statement that an AI tool was used.
Algorithm: a set of rules or instructions that artificial intelligence uses to make decisions, solve problems, or produce answers.
Conversational AI: Generative AI tools that replicate human interaction by using natural language to both receive input and produce responses.
Generative artificial intelligence (AI): A category of web-based tools that use algorithms, data, and statistical models to draw reasonable inferences to create seemingly new, realistic content of its own—such as text, images, and audio—from a set of training data. They are not search engines but rather trained chatbots.
Hallucination: Incorrect, misleading, or nonexistent content produced by generative AI tools. When given a prompt, AI tools can scrape data that exists on a topic and create new content that does not actually exist. The fabricated content may include facts, citations to sources, code, historical events, and other real-world information presented as though it is true, which can make AI hallucinations difficult to identify.
Large Language Model (LLM): Type of artificial intelligence that can recognize and generate text from being trained on vast quantities of datasets. Using neural network technology like transformers, the quality of output is determined by the quality of input. LLMs are fed information until they can interpret and organize accurate responses.
Prompt: Data input like words, phrases, questions, or keywords that users enter to signal the AI tool to generate a response based on those factors. The better the prompt, the better the results. For research AI tools, the prompt is almost always a question, request, or topic posed by the user.
Prompt Engineering or Prompt Crafting: Writing the prompt or input data into the form of a question, request, or topic that will elicit a response from the AI tool. A well-crafted prompt enables the AI to give you meaningful and useful results. A bad prompt may result in irrelevant data or lead you away from the best research.
Transformer: Artificial neural network architecture that mimics human brain functioning and gives AI tools the ability to learn from experience or data without human programming.