This work is licensed under a Creative Commons Attribution 4.0 International License (CC BY 4.0).

You are free to:

  • Share โ€” copy and redistribute the material in any medium or format
  • Adapt โ€” remix, transform, and build upon the material for any purpose, even commercially.

Under the following terms:

  • Attribution โ€” You must give appropriate credit, provide a link to the license, and indicate if changes were made. You may do so in any reasonable manner, but not in any way that suggests the licensor endorses you or your use.

No additional restrictions โ€” You may not apply legal terms or technological measures that legally restrict others from doing anything the license permits.

The Room Got Bigger

AI, Class, and the Uncomfortable Democratization of Understanding

Imagine being thirteen years old and not understanding something your teacher has just explained.

You look around. A few people seem to have gotten it.

You then consider raising your hand, but doing so would put a premature end to the illusion that you’re as intelligent as the kid sitting next to you.

You go home. The kid who looked bored goes home, too. His father happens to have a degree in the subject and enough time between dinner and bed to explain it to him.

You have another homework assignment due tomorrow.

Neither of you is particularly stupid. One of you has access to a human being who can explain something.

Years pass. The intellectual disparity between the two of you becomes increasingly evident, and at some point we decide to attribute it to the individuals themselves, develop metrics by which to quantify their respective intellectual abilities, give one a scholarship, and tell the other that college isn’t necessarily for everyone.

How very equitable.

I’ve been thinking about artificial intelligence in terms of this disparity lately.

For all the talk of the democratization of intelligence, I find myself wondering if we’re already approaching the subject from the wrong angle. Was intelligence itself ever so disproportionately distributed among the socioeconomic classes, or did we simply become adept at mistaking the fruits of intellectual privilege for the intellectual capacities of those fortunate enough to possess it?

There’s a difference.

And several companies, in their pursuit of market dominance, may inadvertently be forcing us to reckon with it.

The Last Luxury Was Having Someone Explain the World

Pierre Bourdieu’s concept of cultural capital seems particularly relevant to the proliferation of artificial intelligence. Wealth affords its recipients more than the physical necessities and luxuries of life. It also affords them familiarity with the language, customs, and expectations of the institutions that govern it.

The child of a lawyer need not become a lawyer to benefit from having one at the dinner table.

There is something almost invisible about this form of privilege.

Knowing how to articulate a concern to a medical professional, challenge the decision of a school administrator, or even determine which government agency you’re supposed to bother with a particular set of problems constitutes a body of knowledge that isn’t evenly distributed. Much of it is acquired through one’s proximity to other people.

And how the hell are you supposed to know what you don’t know?

Education can compensate for some of these disparities. So can friendship, mentorship, and the occasional coworker who takes an inexplicable interest in your future. But the availability of these things is contingent upon circumstances over which the individual has varying degrees of control.

The internet represented a monumental development in this regard. Information that I would have previously had to travel to a library to acquire became available to me in the comfort of my own home. I could spend an afternoon reading about the history of religious thought, learn how to produce music, or acquaint myself with some branch of mathematics that I would inevitably lose interest in the moment the lecturer started writing on the board.

But the availability of information doesn’t necessarily facilitate its assimilation.

I can download a paper on quantum field theory right now. I can also stare at it for forty-five minutes, convince myself I understand it, and become a nuisance to every physicist unfortunate enough to encounter me for the remainder of my life.

Access to information? Achieved.

Comprehension? Not so much.

What I may have needed, rather than another repository of information, was someone capable of meeting me at my particular level of understanding and helping me get to the next one. Someone willing to listen to my stupid analogies, correct them, and explain why my intuitions failed me in the first place.

Those people exist. They make wonderful teachers.

They also have lives, limited time, and bills to pay.

And if the subject in question happens to fall within the purview of a particularly specialized discipline, I might not even know one of them.

Conversational AI has begun to make a relatively inexpensive approximation of this kind of assistance available to a much larger portion of the population.

An approximation.

I see no reason to dispense with the distinction between a machine that can generate an explanation and a human being who has devoted thirty years of his life to a discipline. The former’s ability to convincingly simulate the latter is both its utility and its danger.

But how many of us had access to the latter in the first place?

This is the question I find missing from much of the discourse surrounding AI. We compare the intellectual assistance afforded to us by machines with the intellectual assistance afforded to us by qualified human beings without accounting for the fact that a substantial portion of the population was not regularly afforded much of the latter.

We can’t do away with the socioeconomic disparity by simply pretending that it didn’t exist prior to the advent of artificial intelligence.

The World Became Impossible to Ignore

The internet has also afforded us something a little less pleasant.

I can pour myself a cup of coffee in the morning, sit down, and within a few minutes acquaint myself with an unprecedented amount of human suffering.

I can look at footage of bombings in another country. I can read the grievances of people who inhabit a socioeconomic reality almost completely alien to my own. I can learn that a product I rely on is manufactured under conditions I find morally reprehensible. I can subsequently search for an alternative product, discover that it’s manufactured under similarly reprehensible conditions, and go back to drinking my coffee.

What do I do with all of this?

Ulrich Beck’s conception of a risk society seems increasingly pertinent. Industrial civilization produces risks that transcend many of the geographical and social boundaries upon which its members have relied for security. Its successes generate problems that are increasingly difficult for the beneficiaries of those successes to completely extricate themselves from.

In February of 2026, Iranian retaliatory strikes and debris from intercepted projectiles caused damage to sites in Dubai, including its international airport and the Burj Al Arab hotel. There’s something unsettling about seeing a symbol of extraordinary wealth and apparent security affected by circumstances that its occupants cannot individually resolve.

Of course, the experience of a wealthy traveler whose flight gets canceled isn’t commensurate with that of a family whose home is destroyed. I doubt the former will be joining the latter in a movement for global solidarity on account of his hotel reservation getting screwed up.

But the physical interconnectedness of our world is becoming harder to dismiss. Wealth can mitigate one’s exposure to certain risks without guaranteeing one’s immunity to them.

And the same technologies that allow us to administer, profit from, and otherwise participate in this increasingly interconnected world also furnish us with images of its consequences.

I refer to the disparity between our capacity to witness those consequences and our capacity to meaningfully respond to them as the Global Exposure Paradox.

It isn’t solely a crisis of information overload. I can still inundate myself with literature on the life and times of Immanuel Kant without feeling particularly distraught about my inability to intervene in eighteenth-century Prussia.

What becomes psychologically burdensome is the acquisition of information that is consequential to me, morally or otherwise, in the absence of any apparent means of effecting change.

And even comprehension doesn’t necessarily help.

A 2025 field experiment conducted by Sacha Altay, Emma Hoes, and Magdalena Wojcieszak found that encouraging participants in France and Germany to follow reputable news organizations on Instagram or WhatsApp improved their knowledge of current affairs, their ability to distinguish accurate news from falsehoods, and their trust in news media.

It did not produce a statistically significant improvement in their sense of political efficacy.

A more informed person isn’t necessarily a more empowered person.

Rather, he’ll probably become a more articulate pessimist.

The internet made more of the world available to our perception. It did comparatively little to guarantee our ability to understand or respond to what we perceived. We now inhabit a civilization whose members can regularly bear witness to the consequences of decisions in which they had little or no participation.

And we wonder why everyone seems exhausted.

Who Gets to Ask Why?

This is where I find myself cautiously optimistic about humanity’s future regarding AI. Beyond all of the copyright concerns, Luddite ramblings and fears, etc., there is still hope for those who place value on education and media literacy.

The Reuters Institute’s 2026 research on generative AI found that, across six consistently surveyed countries, weekly use of AI for obtaining information increased from 11 percent in 2024 to 38 percent in 2026.

More interesting to me is what people are doing with the news specifically. Among respondents who used AI for news across the surveyed countries, the most common activity was asking follow-up questions about a story.

Asking follow-up questions.

What a remarkably ordinary thing to have become a technological innovation.

Imagine reading an article about a piece of legislation that directly affects your livelihood. The article makes reference to another piece of legislation, several economic concepts that you’re only vaguely familiar with, and a legal precedent that nobody has bothered to explain.

You can now highlight a paragraph, throw it into a chatbot, and ask what the hell any of it means.

You can ask why the journalist chose to emphasize one particular aspect of the legislation. You can ask for its historical context, inquire as to the motivations of its proponents and opponents, and request the text of the legislation itself.

You can even ask the machine to explain the matter to you again, having understood absolutely nothing the first time around.

Consider what this means for a person with a cognitive disability who has trouble processing lengthy articles, an elderly person who struggles to navigate contemporary websites, or someone trying to make sense of a bureaucratic process in a language other than their native one.

These are distinct problems experienced by distinct people. None of them imply a deficiency in intelligence. They are problems of accessibility, and the means by which we have historically delivered information have not always accounted for them.

We seem to regard the ability to successfully navigate the means by which information is disseminated as a prerequisite for being informed.

Why?

Why should one’s participation in society be contingent upon one’s ability to navigate whatever fresh layer of abstraction a software developer decided to introduce between the individual and his health insurance policy this year?

I suspect part of my enthusiasm for conversational AI stems from the fact that the technology can, at its best, take some of the burden of adaptation off of the individual.

I need not acquaint myself with the peculiarities of a dozen different interfaces to learn the history of a subject. I can begin by communicating in the manner that comes naturally to me.

This is no small thing.

The philosopher Miranda Fricker’s work on epistemic injustice concerns, among other things, the disadvantages people face in making sense of their experiences when the interpretive resources available to them are inadequate. I don’t think a large language model can remedy all of the social conditions responsible for this kind of injustice. A machine trained on the linguistic products of an unjust society is likely to reproduce some of those injustices.

But I also think that giving an individual access to a greater variety of interpretive resources comes with the potential to expand his agency.

Paulo Freire differentiated between an education in which knowledge is deposited into the minds of students and one in which students participate in the process of inquiry itself.

Conversational AI can accommodate both.

I can ask a machine what to think about a subject, or I can use it to develop my ability to think about the subject.

The distinction is everything.

The Machine Between Us and Reality

So we’ve discovered a potential means of mitigating some of the intellectual disparities that have plagued humanity for centuries.

Who owns it?

Sheeeeeeeit.

There’s something characteristically absurd about the prospect of liberating ourselves from historical forms of intellectual dependence by becoming dependent upon the computational infrastructure of a handful of multinational corporations.

And I don’t think we should be particularly flippant about the power these corporations wield.

When Google provides an AI-generated synthesis of information in response to a search, it is helping determine the relationship between the disparate pieces of information it has retrieved, it own interpretation.

The implications are enormous.

How much information about an event can be omitted from a synthesis before the synthesis becomes misleading? How does the machine determine which contextual details are relevant? What constitutes an appropriate degree of attention to opposing interpretations? How much of its response is predicated upon its assessment of what I am capable of understanding, or what I am likely to accept?

Even if we assume the people responsible for developing these systems are uniformly benevolent, we’re left with problems that benevolence alone cannot solve.

Interpretation necessitates selection. Selection necessitates omission.

And a machine doesn’t have to lie to me to facilitate a misunderstanding.

It can furnish me with a series of individually accurate statements, the arrangement of which leaves me with a fundamentally inaccurate understanding of the subject.

Hell, I can do this to myself.

I can cherry-pick studies supportive of a philosophical position I’ve grown attached to and spend three hours developing theoretical justifications for my own bullshit.

I can do it faster with AI, too. The machine can have an entire bibliography ready for me before I’ve finished convincing myself of my own intellectual honesty.

Human beings have been interpreting reality for one another for as long as we’ve been capable of communicating. The point of my concern is the concentration of this particular form of interpretive power and the potentially unprecedented degree of personalization with which that power can be exercised.

Two people can ask a machine about the same event and receive substantially different explanations. There are perfectly innocuous reasons for this. One person might be twelve years old and the other might be a professor of political science.

But what happens when a system begins to cater to our ideological sensibilities as readily as it caters to our educational needs?

The two individuals might get along splendidly with their respective AI assistants while becoming progressively less capable of communicating with one another.

Same world.

Different enclosures.

I have no desire to live in a civilization in which every individual gets his own meticulously curated reality and the corporations responsible for maintaining these realities pat themselves on the back for facilitating human flourishing.

Nevertheless, when I compare this possibility with the world that preceded AI, I struggle to muster the indignation necessary to condemn the technology in its entirety.

Did the twenty-four-hour news cycle furnish us with a particularly equitable means of interpreting world events? Were commercial interests absent from journalism prior to Google entering the picture? Was everybody spending their evenings reading primary sources and discussing the validity of their findings with the neighbors?

Were we at some point collectively enlightened and I just slept through it?

We should demand transparency from AI developers. We should demand that these systems distinguish between what is known, what is inferred, and what remains unknown. We should demand the freedom to interrogate their conclusions and the ability to trace those conclusions back to their sources.

But I want to make those demands in the interest of furthering the intellectual agency that the technology can afford us, rather than in the interest of restoring an idealized version of our past.

The fact that new doors have opened doesn’t negate the necessity of keeping an eye on the people who own the building.

The New Inequality

I keep think about private school kids and public school kids.

The latter now has access to a machine that can explain calculus to him, proofread one of his essays, help him understand a difficult passage in a book, and communicate with him in a manner more conducive to his own learning.

So does the former.

And the former may also have access to more sophisticated models, better hardware, more leisure time, better schools, and a number of human beings capable of correcting the machine when it’s wrong.

There goes the utopia.

But the persistence of inequality doesn’t preclude the possibility of a meaningful improvement in the conditions of those who are disadvantaged by it.

The difference between the old form of inequality and the one currently taking shape may lie partly in the degree of personal agency afforded to the individual at the lower end of the socioeconomic spectrum.

A person who couldn’t afford a lawyer might now be able to familiarize himself with the terminology necessary to have a productive conversation with one. Someone who’s been denied a benefit might be better equipped to understand the reason for the denial and the process by which he can challenge it.

These aren’t particularly glamorous applications of artificial intelligence.

I doubt we’ll see a multimillion-dollar advertising campaign in which an elderly woman successfully appeals an administrative decision and the CEO of a technology company emerges from a cloud of fluorescent vapor to announce the beginning of the next epoch of human evolution.

But I’ll take her successful appeal over the epoch.

The real value of intellectual assistance is ultimately expressed in what it enables human beings to do.

And what I find hopeful about this development is that an individual’s capacity to investigate his own circumstances may become somewhat less contingent upon the willingness of other individuals to investigate them on his behalf.

The difference is subtle, but important.

However, we shouldn’t conflate the individual’s enhanced capacity with an alleviation of society’s obligations to him.

If the process of acquiring public assistance is unnecessarily convoluted, furnishing every applicant with access to AI isn’t a sufficient remedy. The process itself needs to change.

I don’t want us to develop increasingly sophisticated methods by which the individual can navigate increasingly dysfunctional systems and subsequently regard his capacity to navigate them as evidence of the systems’ functionality.

That would be a particularly grotesque form of progress.

There is also the matter of the individual’s ability to recognize when the machine’s assistance has become a hindrance. How does a person who lacks expertise in a subject evaluate an explanation of that subject? To what extent does his newfound capacity for inquiry depend upon the quality of his initial education?

We could see some old disparities diminish while others deepen. An inequality that was previously predicated upon access to information could increasingly be predicated upon access to reliable interpretation, and upon one’s ability to distinguish reliability from the mere appearance of it.

The private school kid might still be better at spotting the bullshit.

But the public school kid now has more opportunities to learn how.

That still counts, no?

More Agency, Less Comfort

I wish I could say that the acquisition of knowledge has made me a happier person.

It has honestly only made me much more annoying.

I think there is a tendency to conflate intellectual liberation with the rather pleasant images we associate with liberation in general.

Freedom should feel good, no?

Empowerment should leave the individual feeling empowered.

But what happens when the intellectual capacities we’re developing afford us a greater awareness of the limitations of our other capacities?

The person who becomes aware of the inner workings of a political institution may find himself more capable of challenging it. He may also discover that the avenues available to him for doing so are less efficacious than he initially believed.

The person who learns about the history of his country may find himself more appreciative of certain aspects of it and considerably less comfortable with others.

And the person who spends enough time interrogating his own philosophical positions may find himself in the unenviable position of having to develop new ones.

God forbid.

We don’t simply accumulate knowledge. The acquisition of knowledge sometimes necessitates the reconfiguration of what we’d previously believed ourselves to know.

I think some of us are witnessing this process play out on a societal scale.

I don’t know to what extent AI is responsible for the discomfort we’re experiencing. Economic hardship, war, political dysfunction, loneliness, and a host of other problems had no trouble making human beings miserable before the advent of ChatGPT.

But I do find myself wondering whether some of the discomfort stems from our developing capacity to recognize and articulate the contradictions that have long constituted our social reality.

People are learning the terminology necessary to address grievances that they were previously unable to adequately communicate. They’re discovering historical precedents for what they might have otherwise interpreted as unique failures on their part.

They’re also learning that other people interpret the same circumstances in entirely different ways.

The subsequent disagreements do not need to indicate that we’ve become less capable of understanding one another than we were before. In some cases, they may indicate that we’re actually encountering one another’s perspectives for the first time.

But conflict is not an end in itself. Suffering doesn’t become morally justifiable on account of its potential epistemic utility. A person can learn from a traumatic experience without the experience itself having been a good thing.

Likewise, an increasingly argumentative society isn’t necessarily an increasingly enlightened one.

What interests me is whether the introduction of new forms of personal agency necessitates a period during which certain formerly stable social arrangements become unstable.

And whether we might be mistaking the instability of some of those arrangements for the deterioration of society itself.

I find the possibility encouraging, though I confess I don’t particularly enjoy contemplating it.

Greater intellectual autonomy entails becoming more comfortable with the absence of certainty.

Here come the goddamn Buddhists again.

The Room Got Bigger

We have more opportunities to encounter the world than any generation prior to ours. We are developing increasingly sophisticated means of interpreting what we encounter. What remains considerably less clear to me is whether we are developing the means by which to collectively respond to it.

Hartmut Rosa’s conception of resonance offers an interesting perspective here. There is a qualitative difference between being connected to the world and existing in a meaningful, responsive relationship with it.

I can spend an entire day acquiring information about other human beings without interacting with a single one.

I can learn about the psychological benefits of community, read a sociologist’s assessment of the deterioration of our communal spaces, contemplate the economic and technological factors responsible for that deterioration, and finally close my laptop having neglected to say hello to the person sitting three feet away from me.

Brilliant.

At what point did all of this intellectual activity become a substitute for my participation in the world I was trying to understand?

Perhaps this constitutes the next great disparity we will have to reckon with.

The internet lowered many of the barriers to witnessing and communicating information.

AI is lowering some of the barriers to acquiring and exercising the interpretive capabilities necessary to make sense of that information.

Who or what will lower the barriers to effecting meaningful change in accordance with what we come to understand?

I don’t expect an artificial intelligence company to furnish us with the entirety of the answer to that question. I’m not even sure it’s a question that lends itself particularly well to technological solutions.

We need one another.

This isn’t to say that we need to abandon our technological creations and rediscover some imagined pastoral existence in which human beings were free of alienation, inequality, or interpersonal strife.

I have no desire to die of an infected tooth in the service of my philosophical convictions.

But I think a civilization whose members are growing more capable of understanding their circumstances needs to develop institutions that are responsive to that understanding.

Otherwise, we’re just becoming increasingly sophisticated spectators to our own lives.

The democratization of interpretive assistance is valuable insofar as it helps expand the agency of the individual. The expansion of individual agency is socially valuable insofar as it doesn’t come at the expense of the agency of others.

How do we get these increasingly empowered individuals to cooperate?

That’s the difficult part. And it has been the difficult part for a long time.

I don’t believe the current proliferation of artificial intelligence guarantees a more equitable society. It could facilitate the development of a civilization in which intellectual assistance is ubiquitous and effective agency is distributed more disproportionately than ever before.

It could also help people recognize that disparity, articulate it, and begin addressing it.

The latter possibility is why I remain optimistic.

For generations, many people lived with the consequences of decisions whose rationale was communicated to them in language they weren’t equipped to interrogate. The opportunity to develop that capacity was itself subject to the socioeconomic forces that governed their lives.

Now more of them can ask why.

Some will ask stupid questions, arrive at horrifying conclusions, and/or become so enamored with their newfound intellectual capacities that they’ll need to be reminded of the existence of other people.

And some will ask questions of our institutions that those institutions should have been made to answer a long time ago.

I hope we don’t confuse the former with the latter.

I hope we don’t allow the difficulty of accommodating new participants in our intellectual and political life to convince us that their participation was a mistake.

The room got bigger and, in some respects, it’s an uglier room than the one I remember inhabiting. The conversations are more difficult, the disagreements more apparent, and there are aspects of its construction that I probably would have preferred to remain ignorant of.

But its ugliness was always there and I just couldn’t see it from where I was sitting.

And perhaps the people who built it could use some help figuring out what to do with it now.

-BU


References and Further Reading