UGS 303: Black Austin Matters - Reddick & Thompson

What is it?

In non-technical terms, generative AI chatbots like ChatGPT are chatbots that have  "learned" how to have a conversation with you or respond to your prompts from a large amount of data. This is called a "large language model." Most people have heard of ChatGPT, which is one such model but there are many more.

Can I use a gen AI chatbot as a student?

It is up to your professor if or how you can use a generative AI chatbot like ChatGPT in a course. Review your syllabus or talk to your professor. 

If you do use it, remember to cite it. Avoid plagiarizing by citing any work produced by any other person or entity, whether it is chatbot output, copying and pasting from a website, paraphrasing or using ideas.

Should I use ChatGPT or something else?

It is up to you. UT has licensed Microsfot CoPilot  so there are fewer privacy concerns  and your interactions with the chatbot are not stored or used to train the chatbot further.

What are the problems with using it to help with research?

• Chatbots  are designed to provide "plausible" responses, not credible responses.As a result, the information it presents can be incorrect. It is up to you to evaluate its credibility, just as with any other source of information.
• The majority of data used to train chatbots is from the open web, not from paywalled content. This means that most of the scholarly information the Libraries subscribes to for scholarly research is not included. UT's licensed version of Microsoft CoPilot is able to link at times to some paywalled content but not all.
• Many chatbots make up/"hallucinate" plausible sounding citations to sources that don't actually exist. 
• Many chatbots were trained on information at least a year old, if not more.

How can it be helpful for research? 

Note: Before doing any of these things, make sure that they are allowed by your professor.

• Use it to start brainstorming a topic or research question. For example, you can ask it to list and describe some topics related to your course content. You can then test how viable the topic is by examining the results of your own searches for information alongside any citations it provides. Ask yourself if there is enough information about it or if there is so much that you need to narrow the topic; and if the information is current enough for your research.
• Use it to brainstorm keywords to research your topic in library search tools. You can then use these search terms to search for information in the library.
• If you are researching a controversial topic, you can ask it for the pros and cons. For example, if you are researching self-driving cars, you could ask it to list and describe the benefits of self-drving cars and the drawbacks or problems with self-driving cars. You can then use this list to find sources about those specific benefits and risks and dig deeper.

Some people don't use AI Chatbots because of ethical considerations. Here are a few of those concerns.

• Bias: The majority of information on the internet is in English and presents a western perspective, which means AI chatbots are learning from a biased data set. These biases are present in chatbots' outputs.
• Environmental impact: Generative AI is incredibly energy intensive and has a huge impact on the environment.
• Privacy: Most chatbots collect and share a lot of data about you they draw from your account information, your IP address and your activity while using the chatbot. Explore your options for opting out or use UT's licensed version of Microsfot CoPilot  to mitigate privacy concerns.
• Copyright: Chatbots are trained on information on the internet, including copyrighted information, with no permission from or payment to the creators. There are a lot of court cases about this as of Fall 2024.
• Unpaid Labor: .Chatbots use the prompts you enter to continue to train their models. This means they are using your work to train their subscription models as well, so many consider this unpaid labor.