Many AI tools are intuitive and easy to use. AI can be integrated in the digital services we use to such a degree that the technology becomes invisible to us. So why bother with AI skills? Do we really need them?

The short answer is yes!

The better you become at understanding and using AI consciously, the more value the technology can create for you and support exactly what you want to learn, create, and achieve.

AI literacy can broadly be understood as the knowledge, skills, and understanding needed to engange with AU effectively. It is not just being able to use the technology. In fact, it is more about how we use it. There is a significant difference between: 

• Knowing how to use AI
• Knowing how to use AI thoughtfully, critically, and purposefully to achieve meaningful outcomes

The article Conceptualizing AI Literacy: An Exploratory Review identifies the following four aspects of AI literacy:

1. Know & understand AI: Know the basic functions of AI and how to use AI applications
2. Use & apply AI: Applying AI knowledge, concepts and applications in different scenarios
3. Evaluate & create AI: Higher-order thinking skills (e.g., evaluate, appraise, predict, design) with AI applications
4. AI ethics: Human-centered considerations (e.g., fairness, accountability, transparency, ethics, safety)

AI literacy can be describes as the ability to use AI consciously and critically. It involves reflecting on how AI is used, critically evaluating AI-generated content, and applying it in ways that support learning.

In the context of studying, it is important to be mindful of how you use AI tools to support your academic work without compromising academic rigour, personal responsibility, or independent thinking.

Whereas we previously relied on search engines to find answers to our questions, many people now turn to AI-powered chatbots. At first glance, these may seem like two ways of achieving the same result: a quick answer. However, there are significant differences, both technologically and in terms of the content they provide.

A traditional search engine finds existing content on the web and presents you with a list of links to, or extracts from, websites that match your search query. This allows you to consider the source and its context. In other words, a search engine helps you find information; it does not generate the information itself.

An AI chatbot works differently. It generates a response using generative AI, which has been trained on vast amounts of data and can produce text, images, and audio based on statistical patterns. The response is not drawn from a single source, and it can therefore be more difficult to assess its reliability and trustworthiness.

The technology is evolving rapidly. Many search engines already incorporate AI-generated answers directly into search results, while chatbots are increasingly providing links and references that allow users to explore topics further and understand the sources on which a response is based.

As a result, the boundaries between search engines and AI chatbots are becoming increasingly blurred. This challenges the ways in which we search for, understand, and evaluate information and knowledge.

Need to know about the technology

• Language models do not comprehend the way humans do. They predict probable words based on statistical patterns and neural networks based on huge amounts of training data. Even though they may provide us with convincing and well-structured answers, they do not think, they have no consciousness or intentions.
• There is no guarantee for correct answers. The models can be wrong, invent information (hallucinate), be biased, and provide you with inconsistent, uniform answers.
  You cannot completely control the language models. E.g. the model may provide you with the answers you would like to hear (sycophancy), adapt the ansewers to your level of knowledge (sandbagging), or come across as manipulative. 
• The values the model is based on can be designed and regulated by the supplier. Consequently, the model itself does not necessarily represent the values stored in its training data.

Nice to know about the technology

• There is often coherence between scaling and performance. Larger models, more training data, and computing power usually leads to a better performance. 
• Language models continue to be a "black box" of sorts. We do not fully comprehend why they provide us with a specific answer, or how they get there.
• Language models can suddenly develop "new skills" and be difficult to predict.
• In some cases, language models can perform better than humans simply because they have access to extremely large amounts of data. 

Before we move on, try or AI control room where you get to process a user prompt as an AI chatbot.

AI-tools have the potential to boost your learning and understanding. With AI literacy as our point of reference, we have summed up a number of points that students should pay attention to when using generative AI in their academic work. 

It does matter how we use AI. The objective is to use AI as a conscious choice when it makes sense to do so, and always to use it critically and with reflection. 

As mentioned in point 7 above, there are several structured prompting methods, e.g. models such as CLEAR (Context, Length, Expectation, Audience, Role). A more presice output is not the only advantage of applying a systematic framework; it becomes an exercise in critical thinking(LINK) in itself. 

When taking a systematic approach to prompting, you effectively force yourself into an academic "mental warm-up". Rather than simply asking for a quick, superficial answer, you must actively draw upon your own knowledge to define the scope and boundaries of the task. By explicitly setting the context or asking the model to adopt a particular theoretical perspective (Role), you create reflective friction at the input stage. In doing so, you take control and establish the academic ground rules before the AI is allowed to generate a response. This is a crucial step towards avoiding premature coherence and maintaining your intellectual ownership of the work.

You can also adopt an iterative approach to prompting, incorporating elements of frameworks such as CLEAR as you go. Consider prompting as a guided dialogue rather than a one-way request. Begin with a basic prompt and then use follow-up questions to adjust the level of detail, add academic context, or introduce reflective friction. For example, you might structure your process in three steps:

1. Begin with a broad prompt: "Provide an overview of the key ideas in the theory of communities of practice."
2. Add context: "Make the overview more specific by relating these ideas directly to [insert a specific case or issue from your course]."
3. Create reflective friction: "Make a critical assessment of the overview you have generated. Which important academic nuances or theoretical limitations are overlooked in this explanation?"

By working in small, iterative steps, you force the model to engage more deeply with the subject matter, while you actively control the academic direction and critically assess the content as you go.