Prompt Engineering

Context is the number one ingredient ChatGPT and similar AI's need to reply with a useful response. When using ChatGPT to help with an assignment, make sure to include all the relevant details from the assignment requirements. Prompt engineering refers to the careful crafting of the input or request made to an AI.

The pneumonic CLEAR will help you remember what good prompt writing needs in general:

1. Concise: Be clear and direct in your prompts. Don't ramble!
2. Logical: Give your directions in an orderly manner. Include in the prompt what format you want the answer to be in.
3. Explicit: Be as specific as possible. Include the context that the AI would reasonably need to know what you're talking about and what the desired outcome is.
4. Adaptive: Adjust your prompt based on what was missing or lacking in the first AI reply
5. Reflective: Pay attention to what prompts led to what AI responses. This will help you write stronger prompts in the future.  

Evaluate the AI as a Tool

When working with AI for research, it’s essential to assess both the technology and the information it produces with a critical eye. Check that information is accurate and well-sourced, free from bias, logically consistent, and not framed with manipulative or overly emotional language. Consider the following:

Purpose — What is the intended function or goal of the tool? 

Credibility — Is the generated or presented information trustworthy? Since generative AI can create original text rather than simply returning search results, cross-checking multiple sources is crucial.

Ethics — Are there any moral or ethical issues connected to use of this tool?

Content Uploads — Does the tool request you to provide existing materials, such as documents or images? If so, could copyright be a concern? Is there an option to prevent your uploads from being incorporated into its training data?

Privacy — What does its privacy policy say? In an educational setting, are there potential FERPA implications? 

Reproducibility — If replicating results matters to your work, can the tool support that need?

Funding — Who finances the tool, and could that financial backing influence the reliability of its results?

Training Data — What sources or datasets were used to train the system, or what databases does it access? Evaluate whether the data is broad enough, includes subscription-only sources like library databases, is sufficiently up-to-date, and whether it contains biases either in content or in algorithm design.

Citations — When references are provided, verify that they are genuine and not fabricated or “hallucinated.” 

Role-Playing

Giving generative AI a role to assume gives the AI more context about the type of response that will help you most. For example, ask the AI to play the role of interviewer for a job you're applying for, adopt the persona of a historical figure, or take a position counter to yours in a debate. Assign it the role of a professor to get feedback with an academic lens on your class notes or essay outline. Remember that if you have access to the real thing (like a professor or Librarian), talking to those experts is likely a better use of your time.

Designing Study Aides

Use AI to create flash cards, quizzes, or summaries of your class notes. It can also create an assignment timeline for a particular project and other time management tools! See our Study Skills Guide for more ideas.

As AI becomes more and more independent, and requires less and less input from human beings, the loss of human control is a major concern. For instance, autonomous war machines could make decisions that could literally wipe out humanity. Human control of AI is already pretty tenuous: current AI models still use English to operate, but experts warn that they could generate their own language, which would vastly decrease transparency of their operation.

The debate over AI autonomy also creates a legal gray area where if AI does do anything immoral or destructive it's not immediately clear with whom the blame lies: with the AI, or with the people that made it?

Tips

• Accountability is also a really good reason to make sure you're citing any AI tools you use in your work. Good research relies on a path that can be retraced from the very first idea to the finished product.
• It's generally a good idea not to enable any settings that let AI use your devices, identity, or otherwise identifiable/traceable information without your timely consent. This means that you are able to decide in the moment to give it permission or not on a case-by-case basis.

Because AI algorithms are created by people and trained on what people say and do, any biases present in their behaviors will transfer to the AI. This causes AI to amplify racist, sexist, homophobic, and otherwise biased or prejudiced stances.

For example, facial recognition algorithms tend to negatively flag people of color more than white faces, or in some cases are unable to read darker faces in dark surroundings. AI programs tend to reinforce sexist bias, like rejecting job applications from women much more than men, or flagging certain jobs as "feminine" (like "nurse," "flight attendant," "secretary," etc.).

AI bias is also a big problem in education, as detailed in this article. For example, falsely flagging essays as AI-generated, especially when written by non-native English speakers.

Tips

• Because AI can echo human bias, you want to be vigilant that it's not giving you flawed or incorrect information based on those biases.
• You also want to make sure that you're not creating bias in your AI tools by phrasing prompts in a way that can only lead to one conclusion.
• AI tools can even show a bias towards you, and start telling you that you're always right or always confirming your ideas, whether they're correct or not. Sometimes this will even lead to people developing a syndrome called "AI psychosis."  Be on the lookout for AI tools always telling you you're right, because the whole point of learning is to sometimes be wrong!

Because you don't have to pay AI or give it benefits like you would a human, companies are very eager to train AI tools to do the work of people who will then be fired. For instance, Duolingo announced that it would be replacing thousands of contractors with AI, a decision that was then walked back (at least a little). Microsoft also recently announced plans to replace human employees with AI.

In many cases, workers have been asked to train their own AI replacements, or worse, have been asked to do seemingly unrelated tasks that were then used to train their AI replacements without their knowledge.

According to the World Economic Forum's "Future of Jobs Report 2025," "trends in AI and information processing technology are expected to create 11 million jobs, while simultaneously displacing 9 million others, more than any other technology trend." This may seem like it's still a net positive, but the report also claims "Robotics and autonomous systems are expected to be the largest net job displacer, with a net decline of 5 million jobs." Autonomous systems almost certainly involve AI, so that's still AI replacing human workers.

Tips

• If you concentrate on forming yourself into a well-rounded person with critical thinking skills, you'll still be in high demand. The "Future of Jobs Report" says, "Analytical thinking remains the most sought-after core skill among employers, with seven out of 10 companies considering it as essential in 2025." Additionally, "creative thinking, resilience, flexibility and agility, along with curiosity and lifelong learning, are also expected to continue to rise in importance over the 2025-2030 period."
• Skills and inherently human traits are also still needed, such as "empathy and active listening," according to the report. There will always be jobs where the human touch can't be replaced.

AI might seem like magic, but using it actually has an enormous impact on the environment. AI data centers require an enormous amount of energy and water. This article in MIT News gives an easily-understandable breakdown:

• "Globally, the electricity consumption of data centers rose to 460 terawatt-hours in 2022. This would have made data centers the 11th largest electricity consumer in the world, between the nations of Saudi Arabia (371 terawatt-hours) and France (463 terawatt-hours), according to the Organization for Economic Co-operation and Development."
• "By 2026, the electricity consumption of data centers is expected to approach 1,050 terawatt-hours (which would bump data centers up to fifth place on the global list, between Japan and Russia)."
• "Each time a model is used... the computing hardware that performs those operations consumes energy. Researchers have estimated that a ChatGPT query consumes about five times more electricity than a simple web search."
• "Chilled water is used to cool a data center by absorbing heat from computing equipment. It has been estimated that, for each kilowatt hour of energy a data center consumes, it would need two liters of water for cooling."

According to the Washington Post (cited in this article, in case you can't access WaPo), writing a single 100-word email with ChatGPT is the equivalent of pouring out a bottle of water. And instead of researching ways to be more environmentally friendly, companies ask citizens of the towns their data centers are in to limit their use of resources, as with Texans being asked to take shorter showers, or with data centers' energy usage being responsible for significantly higher electric bills for ordinary citizens.

Tips

• Unfortunately, there is no way to make zero impact on the environment while still using AI. The only way to avoid it is to stop using AI completely.
• If you still need to use AI, try to use it as little as possible, and minimize your interactions with it (i.e., no pleasantries... every time you greet ChatGPT or say "thank you" is another use of water and electricity).
• Turn off unnecessary AI on your devices and online tools, so that they're not automatically generating an AI query without your knowledge. For instance, to get rid of the automatically generated Google AI answer, type "-ai" into your search. Be aware that this is an imperfect strategy, as it will also exclude any sources with the term AI.

Most major AI tools have committed breaches of privacy, with many of them being trained on things like private healthcare records, Dark Web materials, and people's "private" clouds.

Anything entered into an AI tool is not private, despite these tools' claims. For instance, many ChatGPT users' conversations with the tool were recently found in Google search results, exposing sensitive information. Any nurses or doctors who upload patient health information to an AI tool (an increasingly common occurrence) is breaching the Health Insurance Portability and Accountability Act (HIPAA). And since many AI tools are trained on the whole of the internet, even private or copyrighted words or images are subject to being incorporated into their algorithms.

AI can also be used to steal identities, generate explicit images of someone without their consent (it's happened to Taylor Swift, multiple times), and because most AI tools have security flaws, they can be hacked for nefarious purposes, such as this research team that exploited Gemini's flaws to hack into the physical systems of smart houses.

Tips

• Do not upload anything with personally identifiable information to an AI. Even if the tool claims it won't be stored, it will be.
• It's best to avoid uploading copyrighted materials (song lyrics, articles, books, video clips from TV shows or movies, and more) to AI tools. This includes materials you're uploading for them to summarize.
• Just as detailed in the "Autonomy/Accountability" tab, it's generally a good idea not to enable any settings that let AI use your devices, identity, or otherwise identifiable/traceable information without your timely (AKA right now) consent.