STEP TWO: Think critically about how GAI affects your learning

Will using GAI help or hinder my learning?

There is no one-size-fits-all answer to the question, “Will using GAI help or hinder my learning?”. The answer will depend on the specific task, the specific tool, and how you’re using the tool to accomplish the task. For example, if your goal is to memorize the content of a textbook chapter, asking AI for a podcast or quiz based on the chapter could be helpful. (Asking for a summary is much less helpful and can give you a distorted or incomplete picture of the content.) On the other hand, if your goal is to develop advanced reading skills in your field, you have to read the chapter yourself (but you could use an AI-generated outline or question set to guide your efforts). When deciding whether and how to use GAI tools for academic work, consider the following questions:

• What is the purpose of this task? Academic assignments are designed to help you build the knowledge and skills you need to succeed in your classes and professional life. Do you know why you were assigned this task? Will using a GAI tool support or undermine this goal? If you’re not sure, ask your instructor!
• What specific skills are you outsourcing? The point of academic assignments is not just the product but, more importantly, the learning that happens along the way. Be careful about outsourcing tasks designed to help you build essential cognitive and professional skills. Researchers have demonstrated that relying on GAI may come with cognitive costs, including impact on critical thinking, writing, and coding abilities.
• What specific skills are you gaining? Prompting GAI to generate strong outputs is a skillset in its own right. It includes a number of sub-skills such as (1) analyzing the task’s purpose, scope, and audience; (2) writing prompts that are detailed, logical, and concise; (3) re-prompting to refine outputs; and (4) critically assessing outputs for accuracy and bias. If GAI prompting is a required part of your educational or professional experience, make sure to develop the skills needed to do it right.

Ask Yourself:

• Is this tool helping me learn or just helping me complete the assignment?
• What specific cognitive tasks am I outsourcing and why?
• What important skills would I be developing or strengthening if I did this task myself? What skills would I be developing if I used GAI?

Can I trust GAI outputs?

As a human user, you are responsible for analyzing and fact checking any GAI outputs you incorporate into your academic or professional work. Users with low levels of GAI literacy are the most likely to trust GAI outputs without questioning them, so it’s important to educate yourself (and others) on common GAI pitfalls:

• Illusion of expertise: Because chatbots are designed to deliver outputs that please the human user, they can generate responses that seem correct even when they aren’t. When information is not available in the data used by the GAI tool (for example, in the case of subjects that are highly specialized, new, or not well-represented on English language websites) the algorithm will still generate a reply that sounds plausible at first glance but may be incomplete, incorrect, or biased in ways non-experts (including most students) can easily miss.
• Fabrications (sometimes referred to as confabulations or hallucinations): From time to time, GAI tools might randomly produce nonsensical, inaccurate, or dangerous outputs.
• Bias, overrepresentation, and underrepresentation: GAI models not only replicate and amplify biases found in their training datasets but also reinforce common viewpoints while leaving out less familiar or marginalized perspectives. This is because GAI models tend to average across what they have seen and amplify patterns that appear frequently. As a result, GAI outputs reflect the knowledge and values that are already present in the mainstream (for example, sexist or racist stereotypes), rather than helping you think in unconventional ways or discover less known points of view.
• Too much agreement: Chatbots are engineered to produce outputs that confirm users’ existing opinions to make them feel good, which leads to increased user ratings. By over-relying on GAI tools, you might be missing out on honest formative feedback critical to your growth.

Ask Yourself:

• Am I accepting GAI answers without questioning them?
• Do I have the expertise to assess GAI outputs?
• What stereotypes might be masquerading as facts? Whose perspectives might be overrepresented, underrepresented, or missing from this output?

What are the hidden costs of GAI?

Generative AI tools can perform many tasks much faster and more precisely than a human. But this speed and convenience come at a price. In addition to the labor justice, copyright, and bias concerns discussed in the previous sections, you need to be aware of GAI data privacy concerns and environmental costs. You are responsible for understanding how your data might be used and protecting your own privacy and that of others. You are also responsible for thinking about how your digital choices are affecting the planet. Before defaulting to using a GAI tool, consider the following:

• Data security: When you interact with GAI tools, your inputs may be stored, analyzed, or used to improve future responses. Always check the privacy policies of the tools you use and avoid entering information you wouldn’t want shared or reused.
• Intellectual property: When you upload an essay, article, or book chapter into a chatbot, you might be giving up someone else’s intellectual property, including copyrighted information, without their consent. Before uploading content (your own or someone else’s) make sure you are not inadvertently helping devalue human intellectual and creative labor.
• Environmental costs: GAI tools require massive computing power, which means they consume significant amounts of energy and resources. Large data centers used to train and run these models create enormous environmental strain, including increased carbon emissions and hyper-intensive water consumption. Interacting with a chatbot may feel low-impact, but it’s important to recognize that there is an environmental cost to every prompt and output. For example, generating the “AI summary” for a Google search uses significantly more energy than a traditional search. (To skip the “AI summary” feature, add -ai at the end of your query.) 

Ask Yourself:

• Is using GAI for this task worth the environmental cost?
• Do I know what this tool does with my data?
• Am I sharing someone else’s intellectual property without their consent?

Go to Step Three: Take full responsibility