António Dias de Figueiredo: “If people stop exercising their minds, they may lose their ability to reason”

In ELM’s interview, António Dias de Figueiredo, retired professor of Information Systems at the University of Coimbra in Portugal, points to specific benefits and potential dangers of using AI in adult learning – and offers a gloomy assessment of the prospects for equality in digital learning across Europe.

How does generative AI (GenAI) affect the learning processes and education of adult learners? If learners increasingly use GenAI to write notes, summarise books and analyse information, will it jeopardise or undermine their learning processes?

“It’s helpful to notice that using technological tools may involve three levels of cognition: delegated, augmented and collaborative cognition,” says António Dias de Figueiredo.

“Delegated cognition is when we use a tool to save us the effort of reasoning, when we’re said to be delegating some of our cognitive abilities. This happens, for instance, when we use a calculator,” explains Figueiredo.

This delegation – or “cognitive offloading” – of our cognitive abilities can have both positive and negative consequences, he says.

“On one hand, it can free up our cognitive resources for other tasks, such as creative thinking or problem-solving. On the other hand, it can lead to laziness in thinking and a decline in our cognitive abilities,” says Figueiredo.

“The second, augmented cognition, is when a tool enhances our cognitive abilities, making us smarter for specific purposes. For instance, when we use a GPS app to navigate in a new city, the app frees up our mind for driving. In this sense, it becomes an extension of our own cognitive system.”

According to Figueiredo, the concept can be taken further by the extended mind theory, which postulates that some of the tools we use can become integral parts of our cognitive processes.

“For example,” he says, “a blind person with a cane can develop a better cognition of the world than one without a cane.”

Augmented cognition is when a tool enhances our cognitive abilities, making us smarter for specific purposes.

The third level of cognition is known as collaborative cognition, where humans collaborate with intelligent tools that learn from human input.

“This creates dynamic partnerships to solve problems, make decisions and even engage in creative endeavours,” he says. “When we collaborate with a writing assistant that corrects grammar but also suggests ideas and analyses style, the cognition arises incrementally in the context of the cooperation between human and tool, for example.”

“How GenAI tools affect the learning process depends on the kinds of cognition that are called upon,” says Figueiredo. “If adult learners use these tools wisely as extensions of cognition or as cognitive collaborators, they can improve their cognitive abilities by orders of magnitude. However, if they delegate too much, their learning processes will deteriorate and their cognitive abilities will decline,” he warns.

“Neuroscience shows us that the connections in the human brain are strengthened or reduced depending on how we use them,” he says. “This is the phenomenon of neuronal plasticity, which underlies much of our learning processes. If people stop exercising their minds through reasoning, they may gradually lose their ability to reason.”

Since we can only enhance our understanding and thought processes by reading, pondering and writing by ourselves, is it harmful to let AI do all that?

“Yes! It can be harmful to delegate cognition to tools on a regular basis. When people delegate their ability to do math to calculators, they forget how to process calculations. If they delegate their ability to reflect, they’ll stop being able to reflect beyond the very simple. If this became a common practice, we would have a generation of stupid people in a world of intelligent machines,” the professor says.

If people delegate their ability to reflect, they’ll stop being able to reflect beyond the very simple.

“If only to avoid becoming stupid and prematurely senile, individuals should read, ponder, write, and elaborate texts by themselves, with or without the assistance of AI tools. When they resort to AI tools, they should use them to extend their cognition, and thus deepen their reading, enrich their pondering, improve their writing, and elaborate more deeply on their own ideas.”

What about equality in digital learning? Is it a real possibility for anyone in Europe? Can digital learning offer paths to inclusion, cost savings and more environmentally sustainable learning platforms?

Figueiredo is bluntly pessimistic about digital learning in Europe, noting that it faces a variety of challenges such as skills, access, resources and social interaction.

“While digital learning offers potential for inclusion, cost savings and environmentally sustainable education, it’s only suitable for learners with high discipline and autonomy,” he says. “This substantially narrows its potential. Most learners’ autonomy and self-learning skills have not been sufficiently developed by school systems. Therefore, even if ambitious European programs are launched to promote digital learning, the results are likely to be scanty due to the lack of fertile ground to flourish.”

While digital learning offers potential for inclusion, it’s only suitable for learners with high discipline and autonomy.

Furthermore, he says, access to technology and internet connections is unequal and often problematic across Europe – and doesn’t seem to be improving for the populations that are most affected.

In his view, the quality and accessibility of digital learning resources varies greatly. And though digital learning can be strengthened by bringing in a social dimension, setting up supportive learning communities is not yet a common cultural practice around the region.

What is the situation of marginal groups and minorities in Europe regarding access to the benefits of digital learning?

“I’m not optimistic in this respect either,” Figueiredo responds. “The pandemic has shown that the digital divide is still deep for marginalised groups. Geographical location, poverty and cultural backgrounds all limit access to the internet and other technology.”

He also points out that the diversification of populations associated with immigration is confronted with a lack of digital resources in minority languages, which in his opinion need stronger support.

“AI tools could help in this respect, but teachers in the field aren’t trained and are too overloaded to be able to help. Also, the content and delivery of digital learning materials is often not culturally relevant for the minority groups. Finally, while digital tools can be helpful for some disabilities, a lack of accessibility features and training for educators still excludes many students with specific needs.”