AI Literacy: Reframing Attitudes

This is a thought piece I wrote for a panel on Algorithm Literacy that I was invited to moderate at UNESCO’s Digital Learning Week in September. Unfortunately, I wasn’t able to make it to Paris, but still worth thinking about for 2024, so sharing it here…

Recently, out of curiosity, I asked Chat GPT 3 for its definition of algorithm literacy. Its list of working definitions and main categories was one that many education researchers might agree on: 

• Knowledge of fundamental concepts and techniques
• The acquisition of algorithmic thinking along with the ability to approach tasks and problems in a systematic way.
• The ability to evaluate algorithmic bias and fairness based on an adequate understanding of how algorithms are developed, trained, and deployed.
• The recognition of algorithmic impact, the broader implications of algorithms on various aspects of society, and the ability to analyze and discuss the positive and negative impacts. Finally,
• The ability to engage in discussions about the ethical dimensions of algorithmic decision-making.

Though I did not ask Chat GPT why algorithm literacy was important, it offered its own assertion that “algorithm literacy is crucial in an increasingly digital and data-driven world. It empowers individuals to navigate and participate in a society heavily influenced by algorithms, make informed decisions about algorithmic systems and advocate for fair and responsible algorithmic practices.”

Then I asked Chat GPT to give me a list of research articles that defined algorithm literacy. It gave me ten articles. I will not list them, because it was soon clear that not one of them actually existed. The fact that these articles were fabricated was not something Chat GPT revealed until prodded, though when challenged, it was profusely apologetic for the errors.

Being from the last generation to have lived most of my life in a world without AI, I am not quick to assign authority to a young chatbot that it hasn’t yet earned. For a generation that has never known a world not mediated by AI algorithms, however, this may not be as intuitive. So, our first responsibility, even before we aspire to a consensus on what constitutes AI literacy, should be to maintain and protect the healthy skepticism with which we must all continue to approach AI in the world of education and in our daily lives. 

Any framework that is created to define and support AI literacies is unlikely to take root without also addressing the perspectives and attitudes that are essential to sparking curiosity and building perseverance. 

Building Knowledge 

Chat GPT is right that we live in a society heavily influenced by algorithms. So, it is right for us to call on teachers to prepare students to better understand not just the impact, but also the mechanics of how AI algorithms work [1].  It is important that teachers and students  understand the difference between traditional, or “rules based” algorithms, compared to the more data driven algorithms of machine learning. This does not necessarily mean, however, that a strong knowledge of  traditional programming is an essential building block in developing “next generation” algorithmic or computational thinking [2]. To prepare students and educators for the quickly accelerating impact of AI on their daily lives, we will need to prepare students and teachers to dive right into an understanding of how machine learning algorithms work, even before, or at least side by side with the learning of simple programming algorithms. 

Alongside this call to greater algorithm literacy, it is essential to reinforce other supporting literacies such as data literacy. While data literacy is already in many ways considered a fundamental literacy for science literacy, it continues to be mostly defined in traditional terms, outside the context of AI, as the ability to explore, understand, and communicate with data in a meaningful way [3]. In a world in which our data is increasingly used, and communicated to us by machines, however, we will also need to understand how machine learning algorithms make sense of our behaviours through the data we provide. 

Education systems must do more than simply tack this on as an advanced level of data literacy. If we are to prioritize equity, exploration of the interdependence of humans and intelligent machines is not something that can be limited to specialists, or adults, or to only the most proficient students. Ways to seed how data is used in machine learning can and need to be brought into education plans as early as possible. Fortunately there are devoted and creative researchers who have been working with children as young as seven [4], developing the tools and resources that make it easier for students to at least start to familiarize themselves with and develop some of the big ideas that will spark curiosity and inspire further investigation. 

Building self-efficacy

Through The Algorithm and Data Literacy Project, Digital Moment has provided resources to encourage exploration and eventually the self-efficacy that will be needed if we are to prepare all citizens for the era of AI. We have also highlighted some big ideas that are important to develop alongside the competencies:

• AI is not magic
• Algorithms are not always right
• The data we use to train algorithms matters
• AI comes with ethical issues, like bias

Before we can develop self efficacy among students, however, we need to face the greatest stumbling block to the acquisition of this knowledge, self-efficacy among teachers. What we have seen again and again working in this space is that teachers who do not believe in their own ability to understand how AI works are less likely to believe that students can acquire this knowledge. 

Curriculum such as MIT’s DAily , and Raspberry Pi’s Experience AI have been carefully calibrated to manage the cognitive load of learning new technology and can support teachers in acquiring a belief in their own abilities to conceptualize and experiment with AI in their classrooms. But the right curriculum can’t in itself provide the motivation to learn. When learners believe that content is beyond their ability to understand, and requires technical expertise beyond their capacity to employ, curiosity is snuffed out before it has a chance to ignite.  

Seizing the moment

Motivation is often a function of  belief in our own ability and the right internal and external triggers. One such trigger is clearly the emergence of the latest models of Chat GPT. The impact of AI on the teaching profession is now impossible to ignore. Every year more and more students will be handing in homework that has been either done, influenced, or at least developed through the use of generative AI. As educators and their students will inevitably discover, Chat GPT will be helpful, but often wrong and misleading.

Fortunately teachers can start right here in investigating and exploring the strengths and weaknesses of AI in this moment in time, in scoping the problems that it can and cannot solve, in conceptualizing the mechanics of generative AI, and in sparking and developing the belief that they can and will develop next generation algorithmic and data literacy alongside their students. 

[1] Webb, M.E., Fluck, A., Magenheim, J. et al. Machine learning for human learners: opportunities, issues, tensions and threats. Education Tech Research Dev 69, 2109–2130 (2021).

[2]M Tedre, et al. Teaching Machine Learning in K–12 Classroom: Pedagogical and Technological Trajectories for Artificial Intelligence Education,  IEEE Access vol.9, pp 110558-110572 2021PP(99):1-1  

[3] https://www.statcan.gc.ca/en/wtc/data-literacy

[4] Druga, S. Growing Up with AI: Cognimates: From Coding to Teaching Machines. Doctoral Dissertation, Massachusetts Institute of Technology, Cambridge, MA, USA, 2018