Can AI Support Critical Reading?

Can AI Support Critical Reading?

#1 The Questions That Guide Understanding

Prologue

“Can AI help support – or even improve – our reading habits and our critical understanding of complex texts?”

That is the question I started asking myself a few months ago. I had not been using ChatGPT for a while. My first experiences with OpenAI’s language model had left me both impressed and disappointed when it first launched: too many hallucinations, too many errors – unreliable, in short. I wrote about it at the time, and those who are interested can look it up.

Then, a few months ago, I read about the progress that had been made. I came across NotebookLM, Google’s AI-powered research and annotation assistant, and glimpsed its potential. I started digging further and discovered Anthropic’s Claude, which has been my virtual thinking companion ever since.

To gauge the improvements since those early disappointing experiments, I put these tools through their paces in several ways, which I will describe in this and future posts. I changed my mind. Despite their current limitations, these are tools that can genuinely help and extend not only intellectual work and learning, but also critical thinking. When used well – and this, as I will say often, already requires possessing certain competencies – they can enable greater depth of analysis, understanding, and retention than the strategies we use in the analog world.

What Is This Text Trying to Convince Me Of?

“Before diving into any philosophical text,” I used to tell my students at the start of my philosophy course, “ask yourself three simple questions that will serve you better than any sophisticated hermeneutic theory:”

• What is this text trying to convince me of?

• How does it go about doing that?

• Did it succeed?

This interrogative triad, disarmingly simple as it is, works as a critical key for any informative or argumentative text I encounter – whether it is a philosophical essay, a political manifesto, or even a sophisticated advertisement.

The first question – what is it trying to convince me of? – forces me to identify the central thesis the author wants to plant in my mind.

The second – how does it go about doing that? – invites me to observe the persuasive strategies at work. It is striking how aware we become of rhetorical techniques, narrative choices, and lexical decisions when we actively ask this question. Suddenly, the mechanics of persuasion become visible, like watching a magician reveal his tricks.

The third – did it succeed? – is perhaps the most powerful of all. It gives me back my agency: the power to decide, consciously, whether to accept or reject the proposed thesis, having identified it, understood the strategies used to persuade me, and evaluated the reasons offered in its support.

These three questions work like an X-ray of the text, revealing layers of meaning and intentionality that would otherwise remain invisible. They expose the persuasive architecture underlying even apparently neutral and objective discourse.

This simple interrogative tool is, of course, part of a broader approach to understanding texts. Every argument is built as a response to questions – some explicit, others tacit but no less influential. The ability to see these underlying questions is a privileged point of access to the deeper meaning of what we read.

Today I want to talk about the power of questions. Not the deep, philosophical questions that trigger creative or problem-solving thinking, but the questions that activate the critical mode of thought – questions that set the mind for active comprehension, analysis, and judgment.

The Cognitive Value of Identifying a Text’s Questions

One of the most common and counterproductive ways to use a chatbot is to ask it for a summary of a text, or to identify its key points.

Take this passage from Philippe Ariès’ Centuries of Childhood:

“In medieval society the idea of childhood did not exist; this is not to suggest that children were neglected, forsaken or despised. The idea of childhood is not to be confused with affection for children: it corresponds to an awareness of the particular nature of childhood, that particular nature which distinguishes the child from the adult, even the young adult. In medieval society this awareness was lacking. That is why, as soon as the child could live without the constant solicitude of his mother, his nanny or his cradle-rocker, he belonged to adult society.”

I asked Gemini for a summary, and this was the result:

“Medieval society lacked the concept of ‘childhood’ as a distinct phase of life with its own characteristics, separate from those of adults. This does not mean that children were neglected or mistreated, but rather that there was no awareness of the specific features of childhood that distinguish children from adults. As soon as a child was able to do without constant care, he or she was immediately integrated into adult society, with no further distinctions.”

Now let us change approach and ask the chatbot: “What questions does this text answer?” Here is the result:

• Did the concept of “childhood” as we understand it today exist in medieval society?

• What does the absence of a “sense of childhood” actually mean?

• What was the transition point from being a “child” to belonging to adult society?

• How does medieval adult society appear to contemporary eyes?

• How was this conception reflected in the language of the time?

• In which areas of social life did this lack of age distinction manifest itself?

• How were children represented in medieval visual and social culture?

Even more interesting is asking the chatbot to identify the implicit questions the text is responding to. The result:

1. Is the sense of childhood a cultural universal or a historical construct?

2. Was there a difference between material care and conceptual recognition of childhood?

3. When did the modern conception of childhood emerge?

4. How can we interpret medieval society through our contemporary categories?

5. What practical consequences did this different conception of childhood have?

6. How was this conception reflected in the cultural representations of the time?

7. To what extent does language shape our perception of the phases of life?

8. Was medieval society less “developed” than ours in its understanding of childhood?

9. What is the relationship between physical dependence and social status in medieval childhood?

The summary offers more immediate access to the essential content, while making the questions explicit provides tools for deeper and more contextualized understanding, transforming the reading process from the passive acquisition of information into an active exercise in critical thinking.

Where the summary focuses primarily on explicit content – condensing it while preserving the original argumentative structure – the question-based approach produces a problematization that brings to the surface the fundamental issues the text addresses, including its implicit assumptions.

This difference is not merely formal: it concerns the level of cognitive processing involved. In the first case, we have a synthesis that keeps the reader in a relatively passive position. In the second, a dialogic relationship with the text is established, actively engaging the reader with the questions it raises.

The author’s intentions – such as a critique of presentism or a deconstructive purpose – emerge more clearly through questions, whereas they often remain implicit in the summary.

The question-based approach also encourages connections with other texts addressing similar issues, and opens a wider interpretive space, situating the reading in a critical and dialogic dimension.

It is also worth noting that a chatbot-produced summary risks reproducing the text’s own assertions in abbreviated form, without interrogating their premises. An analysis of questions, by contrast, necessarily leads to examining what the text takes for granted – its implicit assumptions, its blind spots.

Finally, asking a chatbot to identify the questions a text is answering requires the system itself to operate at a deeper level of understanding. Instead of simply identifying and reproducing the salient information, it has to reconstruct the overall problematic that animates the text. This process generates results more likely to stimulate the reader’s independent reflection and critical engagement with the text.

As a demonstration, I asked Claude: “What do these questions tell us about the author’s intentions?” Here is the response:

Analyzing the implicit questions the text addresses reveals several authorial intentions:

Historical-anthropological intent. The questions about the constructed nature of the concept of childhood and its historical evolution suggest that the author intends to offer an anthropological reconstruction of a mentality different from our own. The aim is not simply to describe behaviors, but to reconstruct a conceptually structured historical world.

Deconstructive intent. Questions that challenge the universality of the “sense of childhood” reveal an intention to deconstruct categories we take to be natural. The author wants to show us that what we perceive as obvious and unchanging is in fact a relatively recent historical construct.

Critique of presentism. Questions such as “How can we interpret medieval society through our contemporary categories?” and “Was medieval society less ‘developed’ than ours?” suggest that the author intends to warn against the tendency to judge the past by contemporary standards.

Methodological intent. Questions about language and cultural representations indicate that the author wants to demonstrate the importance of these elements as sources for understanding historical mentalities. There is an implicit reflection on historical method itself.

Comparative intent without value judgment. The text avoids presenting the medieval conception as inferior or primitive. The questions reveal instead an intention to use historical comparison to make the conceptual structures of both periods visible, without establishing a hierarchy of value.

Epistemological intent. Ultimately, the questions reveal an epistemological intention: to invite us to reflect on how our conceptual categories shape our perception of reality, and on how these categories are historically contingent.

The important thing to emphasize is that once the deeper intentions of the text have been brought to light – and they must always be subjected to scrutiny, because the machine occasionally goes off course – a more profound and personal dialogue with the text becomes possible. It is no longer individual assertions that must be accepted or rejected: it is the deeper questions of meaning that lead us to reflect. Take the “comparative intent without value judgment” – does it not open up a whole world of new questions and lines of inquiry?

[embed]How a Text Wins the Reader’s Assent Line of Reasoning and Persuasionpietro-alotto.medium.com

AI as an Archaeologist of Questions

In this game of questions, artificial intelligence reveals a fascinating and underexplored potential: that of archaeologist of hidden questions.

In the course of my research, I have found that AI can analyze a text and identify, with surprising precision, the entire interrogative architecture that sustains it. It can map the invisible structure of the discourse, making explicit what was implicit, bringing to light the interrogative skeleton on which the author built the argument.

What strikes me is that AI systems, by virtue of their ability to process enormous quantities of text, have developed a particular sensitivity to argumentative patterns. They recognize when a paragraph responds to an unstated objection, when an example illustrates a principle that has never been made explicit, when a conclusion presupposes questions the author never openly raised.

Navigating the Persuasive Sea with Awareness

In an era of information overabundance, the ability to map underlying questions is a genuine cognitive superpower. This is not simply an analytical skill: it is a form of resistance to manipulation.

When AI shows me that an apparently objective article on economics implicitly presupposes ideologically oriented questions, I gain a new interpretive freedom. I can ask myself: are these the right questions? Which alternative questions have been excluded? What underlying values are hidden in these questions?

AI becomes not a substitute for critical thinking, but a powerful ally that amplifies the capacity to navigate consciously through the persuasive landscape that surrounds us. It is like having a co-pilot who identifies hidden currents while I keep a firm hand on the wheel.

A Deeper Dialogue with Texts

This practice can radically transform the reading experience. When I have engaged with complex philosophical texts, I have always tried to trace the fundamental questions they were responding to – to understand the author’s intentions, identify the main theses and ideas, follow the logical structure and argumentation, deepen comprehension, and facilitate critical evaluation. A demanding task that used to take considerable time. AI makes all of this much more manageable and faster.

When there is an article that interests me, I no longer simply follow the argumentative flow the author has laid out. I establish an active dialogue with the text, continually asking myself about the questions that structure it.

With AI’s help, I can quickly identify the interrogative architecture of a text and decide whether it deserves sustained attention. I can determine whether the fundamental questions it addresses are relevant to my own inquiry, whether it offers original perspectives, or whether it simply retraces questions already well explored.

At a time when attention is the scarcest resource, this capacity for discernment is invaluable.

A Legitimate Concern

Are we not giving up personal reading by doing this? The unique, unrepeatable gaze that finds meaning in a way no one else can replicate? Are we not at risk of abdicating our critical autonomy – of becoming too dependent? These are legitimate questions. I can only answer from experience: I feel that my reading has deepened, not diminished. A deep understanding gained in collaboration with AI still leaves me fully responsible for judging, evaluating, accepting, or rejecting. That responsibility does not transfer.

A New Interpretive Alliance

What I find most compelling in these experiments is the complementarity that emerges between artificial intelligence and human sensibility. AI excels at identifying argumentative patterns and recurring interrogative structures, while I bring the contextual depth, value awareness, and critical judgment needed for substantive evaluation – which, in the end, remains entirely my own.

This interpretive alliance represents, to my mind, a promising model for integrating computational capacity with human intelligence: not substitution, but mutual enhancement that enriches the hermeneutic experience.

In my personal journey of discovery, I continue to find in this practice an inexhaustible source of intellectual wonder. Every text becomes a universe of questions to explore, a landscape of inquiry to map and traverse with renewed awareness.

I often ask myself: what invisible questions structure the texts that shape your everyday decisions? Learning to make them visible might transform not only how you read, but how you think.