Advanced Listening Practice

Leah: Marcus, I want to start with a question that sounds simple but probably isn't. When we talk about privacy today, are we even talking about the same thing that the same word meant twenty years ago?

Marcus: No. And I think that's one of the central problems, that we're using a twentieth-century concept to describe a twenty-first-century condition. Twenty years ago, privacy was primarily spatial and informational. You had a right to be left alone in your home, and you had some control over who accessed your personal records. Both of those were imperfect protections, but the framework made intuitive sense. Privacy was about boundaries, and you could point to where those boundaries were.

What's happened since then is that the concept has been hollowed out from the inside. The boundaries still exist in law, more or less, but they've become functionally meaningless because the mechanisms of data collection operate at a scale and a level of granularity that the legal framework was never designed to address. Your phone logs your location continuously. Your browsing history is aggregated and sold by data brokers you've never heard of. Your purchasing patterns, your social connections, your reading habits, the pace at which you scroll past certain content versus the content you linger on, all of this is captured, analyzed, and monetized, usually without your meaningful awareness. Privacy hasn't been abolished. It's been made irrelevant by a system that routes around it.

Leah: But the counterargument you hear constantly is that people consent to this. They agree to the terms of service. They click "accept." Isn't there an element of choice here?

Marcus: There is a formal element of choice, yes. And that formality is precisely what makes the system so effective. Let me explain what I mean. A terms of service agreement is a legal document. It is written by lawyers. It is designed to be comprehensive, which means it is also designed to be unreadable, and I don't mean that metaphorically. Researchers have measured the reading level required to understand the average terms of service, and it's consistently above the reading level of the average adult in this country. The median American would need a postgraduate reading comprehension to fully understand what they're agreeing to when they download a free app.

So yes, people click "accept." But calling that consent in any meaningful sense requires you to define consent as the physical act of pressing a button, stripped of any requirement for understanding. If a doctor handed you a forty-page medical consent form in a language you couldn't fully read and said "sign here or you can't see the doctor," we wouldn't call that informed consent. We'd call it coercion dressed up in paperwork. The digital version is structurally identical. The choice is technically present and practically absent.

Leah: Let's talk about what's done with the data once it's collected. You've written a lot about predictive systems. What are the most concerning applications?

Marcus: I'd point to two areas where I think the consequences are most tangible and most poorly understood by the public. The first is predictive policing, which is the use of algorithmic models to forecast where crime is likely to occur and, in some implementations, who is likely to commit it. The pitch is seductive. Feed the algorithm enough historical data, arrest records, incident reports, geographic patterns, and it will identify hotspots before crimes happen. The problem is that historical crime data is not a neutral record of where crime occurs. It's a record of where policing was directed. If a neighborhood was over-policed for decades, the data will reflect more arrests in that neighborhood, not because more crime occurred there in absolute terms, but because more enforcement resources were deployed there. The algorithm ingests that bias and reproduces it as prediction, which then justifies further concentration of policing in the same areas. The feedback loop is self-reinforcing and almost invisible to the people operating the system, because the outputs look like objective analysis. The machine said this neighborhood is high-risk. How can you argue with a machine?

Leah: And the second area?

Marcus: Advertising and content targeting, which sounds less dramatic than policing but affects far more people in their daily experience. The advertising model of the internet, which is the dominant economic model of virtually every free platform, depends on knowing as much about you as possible so that the right ad can be placed in front of you at the right moment. That's the basic mechanic. But what's developed on top of that basic mechanic is a sophisticated system for modeling and manipulating human attention and behavior.

The platforms don't just know what you've bought or what you've searched for. They model your emotional state. They know, based on your behavior patterns, when you're likely to be bored, anxious, lonely, or impulsive. And the advertising is calibrated to exploit those states. A gambling ad served to someone the algorithm has identified as being in a period of financial stress. A diet product shown to someone whose browsing patterns suggest body image anxiety. These are not hypothetical examples. They are documented practices. The system doesn't need to understand you in a human sense. It just needs to predict your behavior accurately enough to influence it, and it does.

Leah: You're describing a system that operates on probabilities. How accurate are those probabilities, really?

Marcus: Individually, not all that accurate. This is something that gets misunderstood. The algorithm doesn't know you. It doesn't need to. It operates on populations. If a model can predict with sixty-five percent accuracy that a person with profile X will click on ad Y, that's more than sufficient at scale. You're not trying to manipulate one person. You're trying to shift the aggregate behavior of millions of people by a few percentage points, and the economics of that are enormously profitable even at modest accuracy levels. The individual experience of being wrong, of seeing an ad that makes no sense to you, is the system misfiring. You barely notice it. The system succeeding is far harder to detect, because when it works, the ad feels relevant. The content feels chosen. You think you found it. That's by design.

Leah: Is there a meaningful reform possible within the current framework, or does the entire business model need to change?

Marcus: I think the business model needs to change, and I'll explain why incremental reform tends to fail. Every regulatory intervention so far, and there have been several, GDPR in Europe, various state-level privacy laws here, has operated within the same basic structure. The assumption is that if we give users more control, more transparency, more opt-out mechanisms, the system becomes acceptable. But the asymmetry of the relationship is so vast that individual control is essentially theater. You can opt out of tracking on one app. You cannot opt out of a data ecosystem that operates across every platform, every device, and every service you use. The data flows are too interconnected and too opaque for individual action to make a structural difference.

What I'd argue for, and I'm aware this is politically difficult, is a shift from an opt-out model to an opt-in model. The default should be that your data is not collected, not shared, not sold. If a company wants access to your behavioral data, they should have to ask for it in plain language, explain specifically what they'll do with it, and offer the service without data collection as a genuine alternative, not a degraded one. That would fundamentally alter the economics of the internet, which is exactly why it's resisted so strenuously.

Leah: You mentioned opacity earlier. One of the things that seems to concern you most is not just what these systems do but the fact that we can't see how they do it.

Marcus: That's exactly right, and I think it's the dimension of this problem that receives the least attention relative to its importance. The algorithms that determine what you see, what you're offered, what price you're shown, what content is promoted or suppressed, those algorithms are proprietary. They are trade secrets. No outside researcher, no regulator, no elected official can examine the decision-making logic of a system that affects billions of people daily. We are governed, in a functional sense, by processes we cannot inspect.

And when I use the word governed, I mean it quite precisely. These systems shape access to information, which shapes public opinion, which shapes elections, which shapes policy. The idea that they are merely private business tools operating in a neutral marketplace is, at this point, a fiction that benefits the people who operate them and disadvantages everyone else. We require transparency from public institutions because we understand that power without oversight is dangerous. We have not yet applied that principle to private institutions whose power over daily life now exceeds that of most government agencies. That inconsistency will, I think, come to be seen as one of the defining governance failures of this era.

Leah: Last question. Are you optimistic that any of this changes?

Marcus: Optimistic is a strong word. I'd say I think change is possible but not inevitable, and the window is narrower than people assume. Every year that passes without meaningful structural reform is a year in which the surveillance infrastructure becomes more deeply embedded, more difficult to dismantle, and more normalized in public consciousness. The most effective weapon these systems have is not their technical sophistication. It's the gradual erosion of the expectation that things could be otherwise. When people stop being surprised by what the algorithm knows about them, when they stop finding it strange that an ad anticipated their thoughts, the fight is largely over. And I think we're closer to that point than most people realize.

Leah: Marcus, thank you. This has been genuinely sobering.

Marcus: Thank you, Leah.