Is prompting so important? Other keys and dimensions of AI literacy

The debate on what artificial intelligence literacy should look like has a minimum normative anchor in Article 4 of the European Union's AI Regulation. This article states that providers and those responsible for the implementation of AI systems must ensure a sufficient level of AI literacy of their staff, based on the type of organization, the level of initial knowledge and the work context.

Neither the text of the article nor the explanation of the AESIA go into detail about specific skills such as prompting, nor do they place it as a central element of AI learning. However, a large number of courses and trainings have placed this ability to write requests at the center of the training journey to learn how to use AI. Is it really differential to know how to structure a request or to know what type of prompt we need at any given time? What other angles can we explore to measure literacy and maturity in the use of AI?

A 360º view of AI literacy

In practice, the quality of our requests depends on four fundamental factors that influence our prompting and the results we obtain, and that have to do, mainly, with the prior knowledge that the person who formulates them has. Building and solidifying that knowledge base should be the fundamental objective of all AI training.

1. Knowledge of the world and the profession

The ability to make effective prompting improves naturally when there is experience in a specific field. People who have worked for a long time in a specific field often formulate better instructions simply because they are able to better identify relevant information and how to arrive at a meaningful result. We tend to think of text generation first, but it is easier to observe this in image or video generation, where professionals specializing in design, illustration or cinema obtain more accurate results because they know the expression codes to obtain what they are looking for: lighting, contrast, planes, coherence or movement. Both our professional maturity and the regular use of AI-based technology will be useful when it comes to autonomously detecting the AI slop, both in written text and in images or videos.

2. Theoretical knowledge of technology

Currently, and especially in the business world, the great trend is to offer eminently practical training and very focused on daily use, even in workshop format, which favors functional training in tools and prompting. However, no matter how practical the approach, basic technical issues should never be left out of training. The operation of machine learning, neural networks, the construction of the internet's large text-based language models, the technology behind predictive models or the layers of rules that are added to digital products are concepts that later help to adjust expectations, make better use of technology and make more conscious use of it.

3. Knowledge of the social and cultural context

Artificial intelligence does not operate in a vacuum: it is conditioned by who configures and trains it, with what type of data, from where in the world and in what language. This leads to linguistic biases, especially the predominance of English and Western Anglo-Saxon culture, which can influence how the world is represented. Biases of race, gender and other axes of inequality are also manifested that must be controlled from use. We must add the business, geopolitical and cultural issues that we need to know to understand the context of AI, the priority of ideas in their answers, what makes sense to ask and what not, and how it can influence our creativity and our thinking. Finally, we must be aware of the state of regulation in Spain and Europe to understand what we can do with the content we obtain from an AI-based system.

4. The role of critical thinking

In the three previous points we have talked about the knowledge necessary to consciously use AI systems, following the first of the pillars of critical thinking: we cannot question something if we cannot relate it to external information. Once this first layer is established, forming the criteria requires cognitive analytical skills, willingness and critical attitude, and a habit of metacognition that allows us to think about how we think and exercise a conscious audit of our use of AI. 

Critical thinking does not arise spontaneously in the interaction with AI, but is built along the life path and from the contrast with information obtained in contexts external to it.

Four guidelines to give meaning to prompting

Taking into account these factors, here are several guidelines that help you make your prompts more effective:

1. To avoid the predominance of English and the Anglo-Saxon cultural framework: indicate in which language or context you are working.

2. To avoid answers that automatically focus on Europe or the US and to expand representativeness: place the geographical and cultural context from the beginning.

3. To avoid implicit interpretive frameworks and force a more conscious reading: clarify from what point of view you want the answer.

4. To avoid conditioning the answer with prior judgments or biases incorporated into the formulation itself: avoid introducing prior opinions or assumptions into the question.

Example:

🔴Prompt 1: "Explain youth democratic participation in countries such as Spain, Sweden and Japan and compare them with each other."

🟢Prompt 2: "Explain the democratic participation of young people in Spain, Sweden and Japan, taking into account the geographical and cultural context of each case. It uses language adapted to an audience in Spain, but avoids assuming that Western models are the main norm or point of reference. It analyzes the phenomenon from different perspectives without taking for granted a single interpretation or cause, and without introducing prior judgments."

Prompting  is an interesting and useful skill that, beyond certain tricks and non-obvious functionalities, can be acquired intuitively, even by imitation and in a short time, once we have the necessary knowledge about how AI systems work. Its prominence today is more due to the fact that it is a competence that is easy to identify and explain, which becomes a training product very quickly. However, if we want to apply a broader and more contextual approach to AI literacy,  prompting must not displace technical, regulatory and cultural content that constitutes the true basis of learning.

Content created by Carmen Torrijos, an expert in AI applied to language and communication. The content and views expressed in this publication are the sole responsibility of the author.