Skip to Main Content

ChatGPT and other Generative AI

When Can/Should I Use Generative AI?

Unfortunately, in many cases there's no easy answer to this question unless the person for whom you're performing the task has given you explicit instructions about what AI use is or isn't acceptable. If you're unsure, it's best to ask them.

Here are some additional considerations for both students and instructors that may help in determining acceptable uses of AI in given contexts:

  • What skills does the assignment ask you to demonstrate? If you used AI, would it be doing the work of demonstrating those skills for you?
  • AI reflects many types of cultural bias. If you are asking it to generate content for you, are you aware of the types of bias you may be replicating?
  • If you are uploading materials (e.g. scholarly articles or student papers) to an AI program, are those materials copyrighted? Are you violating data privacy principles and/or FERPA?

For university-wide guidelines, please consult Wesleyan's Generative AI Usage Policy here.

Bias in AI

Because generative AI draws on existing information on the internet, it inevitably reflects the biases and inaccuracies present in that information. For example, Wikipedia was among ChatGPT's training sources, and Wikipedia's demographic biases are well-documented. These shortcomings are particularly easy to see in image generation. The image below is ChatGPT 4's response to the prompt "Create an image of several female friends hanging out together."

ChatGPT-generated image of seven thin, feminine-looking seated in a living room.

Note that all of the women pictured appear young, thin, and conventionally feminine. (Note also that the woman perched on the arm of the couch appears to have three legs, while the arm of the woman on the left appears to be melded onto her leg.)

Additionally, it's often difficult to know for certain where generative AI is pulling its information from, or whether the information was collected with its creator's consent.

Even if your instructor permits the use of AI, it is important to attribute credit for the work it's done on your behalf. For more information on citing generative AI tools, see the "How to Cite GenAI" section.

Copyright and Privacy Issues

Some AI models were trained on copyrighted data without the permission of the creators of the data. A number of copyright infringement lawsuits have been filed against AI platforms such as Stable Diffusion and OpenAI. Further, uploading material such as academic articles to AI (in order to ask it for a summary, for example), may be a violation of copyright.

While current copyright law does not recognize AI as an author per se, it is typically still important to acknowledge where and how you have used AI in performing a given task.

Many AI tools ingest, store, and use both the questions you ask it and information you put into it, so avoid including private, sensitive, or confidential information in your interactions with AI.

Recently, some academic publishers have begun licensing their scholarly content for use as training data for GenAI tools. While it is too soon to measure the impact of these deals on the content’s intellectual and economic value, ITHAKA S+R has launched a Generative AI Licensing Agreement Tracker to help keep authors and GenAI users informed.