CASC 142.2: Strategies for Academic Success: Using AI to support and structure your work

In our session (and in the slides linked on the homepage), we'll go a bit more into what Large Language Models (LLM) like Chat GPT are, how they work, and how that shapes how we can use them. Consider reviewing that information to enhance your knowledge and allow for more reflective and informed use of LLMs.

Whenever we ask a machine for information and support, we should ask questions about how it works, the limits of what it 'knows', and our personal beliefs and comfort levels with different use cases. Some uses might play to the strengths of the tool more than others, some uses could be damaging to us.

Prompt Engineering Tips:

Adapted from sources linked below

• Focus less on a 'problem' and more on the prompt/creation: AI tools can't empathize, creative problem solve, or identify a path forward. If you are seeking to use a LLM to address a problem or question you are facing, consider what the AI can create for you to address that challenge. 
  • Bad prompt: "Hey ChatGPT, my boss wants me to create tiktoks but they all seem lame and pointless and I very much feel like that old 'hello fellow kids' joke What should I do?"
  • Better prompt, summarize: "Please give me an in-depth summary of research on how brands use tiktok to promote engagement with 16-25 year olds"
  • Better prompt, provide examples humans can use: "Please give me 10 examples of popular tik tok trends from this month"
• Provide Context, including a role and goal: Making up a story (or sharing a story/context) will help LLMs be more specific in the language, considerations, and resources they are analyzing and using from their training data. Considering the 'persona' and 'purpose' (sometimes called 'role') you want the LLM to adapt can provide more helpful results.
  • Bad prompt: What should I know about zebra mussels?
  • Better prompt: You are an experienced and personable environmental educator working near Lake Erie. Explain to an audience of non expert adults what zebra mussels are, how they arrived at the great lakes, the damage they can do to great lake ecosystems (include specific indicator species), current prevention and remediation efforts, and what an everyday person might do to prevent or lessen the spread of this species. 
• Be specific for the type of output and format you are looking for: consider the end product that you want from the tool you are using and describe the format you are looking for.
  • Bad example: tell me jokes about summertime
  • Better example: Write 20 potential jokes related any of the following topics: ''' summer, the beach, vacation, hot weather '''. These jokes should be in a Question and answer format and be short enough to be printed on popsicle sticks.
• Experiment and try again!

AI is not human work:

That's probably pretty obvious. Humans can work with and try to encourage AI to create something, but creating an AI prompt is different from creating something. Consider how you might use AI to support your work, creativity, and growth, not as a replacement for it. 

Learning how to use AI tools can support your ability to conceptualize and use language to describe an output. These skills may be separate from the learning goals for your classes and assignments.

AI uses resources:

Data processing and energy use require resources like minerals, water, and human labor. AI tools like chat GPT have been noted to be particularly resource-intensive, with the average query requiring 500 ml of water (about 17 ounces), compared to a 2010 report that put a google search at .5 ml (about .1 of a teaspoon).

If you choose to actively use AI tools, carefully reflecting on and constructing your AI queries can make it more likely that you can get what you need from these tools with less resource use.

AI includes, and can amplify biases:

AI tools are trained on data and information collected and made available. As a result, these tools can continue and even expand existing biases, especially if we believe them to be 'impartial'.

Knowing the many ways that bias and exclusion can impact AI may help you consider how and in what instances you feel comfortable using and learning from these tools.

AI is private property that profits from use:

There is a reason that the saying “If you’re not paying for the product, you are the product.” has become common in the past few decades. 

AI tools add your queries to their training data and sometimes to a individual profile of what the tool knows about you and your preferences. Carefully consider what you share with these tools.