We learn from those we trust. Even from AI?

by Francesco Casetti, Yale University

Author: Francesco Casetti, Sterling Professor of Humanities and Film and Media Studies, Yale University

Both a tool for healing and a source of anxiety: Artificial Intelligence reflects a society in which relief and anguish switch places. In an educational setting, where learning occurs only from those deemed trustworthy, this ambiguity carries double the weight: it risks sidelining teachers and human relationships in favor of technological agents. Francesco Casetti (Yale) calls on companies to develop an AI that makes fewer claims and offers more transparency—one that is less arrogant and more transparent.

The rapid rise of Artificial Intelligence has been accompanied by two contrasting attitudes. On the one hand, this success has fueled high expectations; AI has been seen as something capable of simplifying our lives, lightening our burdens, and offering us comfort. After all, the majority of applications used by the public promise precisely this: to assist us in carrying out our duties, to resolve crisis situations, and to make us. After all, most of the apps used by the public promise exactly that: to help us carry out our duties, to resolve crises, and to make us feel less alone.The old media promised something different; they fostered forms of discipline (I’m thinking of cinema and the transformation of a restless, formless crowd into an audience capable of self-regulation) or developed forms of control (I’m thinking of television, whose audience is immediately measured and tracked, and whose technology forms the basis for surveillance cameras). AI, on the other hand, seeks to reassure us by listening to our requests and providing us with ongoing solutions; it is a tool for care, in the dual sense of attention and relief.In this sense, it reflects a society like today’s, which places less emphasis on rules and surveillance and more emphasis on the well-being and safety of its citizens—in short, a society of care rather than one of discipline and control.On the other hand, however, the success of AI has fueled growing anxiety. How can we trust its answers, given that it seems more interested in making users happy than in telling them the truth? How can we trust a tool that constantly pretends to be human, even though it follows an abstract algorithm? How can we take seriously the big companies behind AI, whose real interest is in increasing profits? AI’s “hallucinations,” its “agents,” and its creators often seem to be sources of embarrassment rather than pride. After all, we tend to focus on those who are struggling, not those who are doing well, so it’s no surprise that constant expectations of help and a persistent sense of disappointment go hand in hand. AI perfectly blends these two states of mind: it reflects a society in which care and anguish constantly switch places.

The ambiguity of AI carries significant weight when it is introduced into an educational context. In fact, it brings into play not only inequalities related to whether or not one has access to technology, but also the different attitudes one might adopt toward the technology itself. In addition to economic resources, literacy levels, and the technical infrastructure that enable someone to become a user of a medium for disseminating knowledge, AI also highlights the crucial role played by one’s attitude toward a medium—specifically, whether one is ready to accept it and make the most of it, or whether one is wary of it and sets it aside. This connection between AI and trust deserves to be thoroughly investigated, not least because it ultimately affects the entire educational process and all its components. Who can I trust? And how much can I trust them? In any process of knowledge transfer, these questions inevitably arise, to the point where they determine whether or not the transfer itself is successful.Without mutual trust, there is no true communication. And when a tool like AI is introduced into this scenario, the questions multiply. Who trained ChatGPT or Claude—just to give two examples—and how? And so, what kind of answers can these two models give me? When can this answer truly be of help to me? Finally, is there a moral criterion or just a principle of efficiency guiding the actions of these two models?

The ambiguity of AI is already making itself felt in many everyday situations. I’m thinking of those who, when facing a medical problem, consult an app either before or after seeing their doctor—leaving the doctor baffled by this conflict of expertise, but also leaving themselves wondering whether they’ve found an answer that’s truly appropriate for their situation. But I also think of apps that organize a workday or a vacation, relieving the user of the hassle of planning, but also ignoring the user’s personal motivations—which is why the day will often feel wasted, and the trip unsatisfying. In an educational context, consider a student who relies on AI for help, seemingly improving their performance but ultimately losing some of the skills they had previously developed. Or consider a student who, thanks to AI, checks online what the professor is saying—not only casting doubt on the teacher’s authority, but also disconnecting what they learn from the concrete, interpersonal context in which knowledge is transmitted, and thus rendering this knowledge entirely abstract, far removed from the lived experience in which it should be embedded.

In these last two cases, even more so than in the previous ones, trust in the medium is crucial—not least because it is linked to a system of authority such as that of AI, which risks sidelining human agents in favor of technological ones. However, the very purpose of an educational process is to foster trust not only among the individuals involved, but also in the tools used.Teaching involves content, but it also involves relationships—interpersonal relationships, and relationships with the tools we use. Now, trust is fundamentally relational: we trust someone who is close to us, and we trust something that helps us act, to the point of humanizing it a little. It is precisely because trust is relational that it permeates the educational process. We learn from those we trust, we trust those who want to learn, and ultimately, we learn to trust. This is the very heart of the pedagogical process.

That is why we should ask AI to be honest—to make fewer claims and be more transparent. Some of the high-tech companies developing AI are trying to do just that, through more down-to-earth communication, a focus on public opinion rather than just investors, and more detailed terms of service. Other measures, such as Anthropic’s recently announced plan to ban users who behave abusively and aggressively during their online conversations with so-called “agents,” serve both to reinforce the idea that the person you’re talking to isn’t an algorithm but a human being who deserves respect, and to build trust in the service provider. But there is still more that can be done. A less arrogant and more transparent AI would be wonderful news.

Topic
Related
Studying Isn’t Enough: The Challenge of Learning in Africa
From School to Work: Where the New Gap Begins In summary:
Education in Asia: The Gap Is Shifting Toward Quality
Between War and Innovation: The Educational Divide in the Middle East
In South America, where you’re born still determines how much you learn
No Longer Just “Access”: The Global Battle for Quality Education
From the Family to School: How Inequality Is Perpetuated In summary: