Using AI

As we reviewed in this unit's introduction, AI holds great promise, but also the potential for great misuse. As students pursuing higher education, it might be tempting to let AI do your work for you, but there are several limitations and drawbacks to using AI for your coursework.

Most importantly, you will not develop your ability to think critically and innovatively - skills that employers are seeking in virtually all fields. In addition, AI is a new technology, and if you don't use it effectively or ethically, you may submit work with the following limitations and errors.

Accuracy & “Hallucinations”

Can you trust that the information you receive from AI-based tools is correct? No.

Inaccurate information may be due to a lack of information in the large language model's "corpus" or body of knowledge. However, some chatbots (like ChatGPT) were designed to produce content that seems realistic, producing inaccurate content with the same level of confidence as accurate content. In fact, it is well-known that AI chatbots will even make up references to appear credible.

It's up to you to determine which is which. You'll need to employ a variety of strategies to double-check information before assuming it is correct. You'll learn more about how to do this in the following section.

You may also want to see what you can learn about the training materials for the different AI tools you are using. For example - some research AI tools pull only from peer-reviewed journal articles, while others pull content more broadly. Because of this, they may be using information that is false, and/or biased.

Bias

While it may be tempting to think of output from an AI-based tool as neutral like digital 1's and 0's from a computer, but it is not. Machine learning models are trained on real-world datasets, and since the world contains misinformation and bias, outputs from these models may replicate or even exacerbate biases we see in the world around us.

Learn about bias in AI: the gender shades video reviews a study of how well popular face-recognition software identifies people of different genders and skin types.

User Error

To get the most out of AI chatbots, you will need to learn how to "engineer" prompts. Trying to get quality output from a chatbot using the search strings like you do in a Google search won't be too effective. Review this guide from Georgetown University on developing AI prompts to learn more.

Copyright Concerns

Where does the content come from? Because machine learning takes huge inputs of data sets, many models use information from the internet in their training. Artists and authors have criticized AI-based tools for using their work without compensation or credit. If an AI-based image generator can produce work in the style of a certain artist, should that be seen as stealing, or paying homage?

Citing AI

AI Citation

If you're using AI to generate text or images, you should acknowledge that in your citations, just like you would for a human author, artist, or photographer. If using AI is allowed on an assignment your professor gives you, each time you include something AI-generated, you have to cite it. (Remember that you can only use AI on an assignment if your professor specifically says so, and always double-check with your instructor if you have questions about AI usage or citations.)

APA Citation Components

APA Style blog: How to cite ChatGPT

Format

Company that made the tool (date text was generated). AI tool (version of tool) [Large language model]. URL.

Examples

In-text citation: (OpenAI, 2023)
Reference: OpenAI. (2023). ChatGPT (Mar 14 version) [Large language model]. ChatGPT Example Conversation.

References:

Adapted from content created by the University of Maryland under a Creative Commons Attribution-NonCommercial 4.0 International License