DL 440

The Veil of Sincerity

Published: July 22, 2026 • 📧 Newsletter

Hi all, welcome back to Digitally Literate.

This week I've been spending a lot of time rebuilding the look and feel, or the aesthetic of my main site and newsletter. As I do this work, I'm constantly reminded of the knee-jerk reaction we have these days. If something looks too polished or too well thought out, it must be AI.

Years ago I described this as the veil of sincerity. The surface of a text signals that it can be trusted, and we judge the veil rather than what's behind it. What's changed is the direction. Polish used to signal care. For now, it signals a machine. The judgment still runs on appearance either way. Sometimes the surface hides work that never happened, sometimes it hides work that did.

Three stories this week about mistaking the appearance of something for the thing itself.

As always, your support is valued. Reach out anytime at hello@wiobyrne.com, and subscribe if you haven't already.

🤷 The Disappearance of "I Don't Know"

For years, we've worried that AI might give us wrong answers. This week's stories suggest a different problem: it may be changing our relationship with judgment itself.

Findings from a recent study suggest that use of AI advice collapses people's willingness to admit uncertainty. Across five experiments with 3,132 participants, merely having access to AI nearly eliminated participants' willingness to suspend judgment. When given access to AI advice, people's willingness to say "I don't know" fell from 44% to 3%. Their accuracy on the same questions fell from 27% to 9%. Their confidence rose from 30% to 76%.

Brown economist Roberto Serrano saw the same pattern play out at institutional scale. After a campus shooting left students traumatized, he replaced his usual in-person midterm with a take-home exam. Forty of 86 students earned perfect scores, pushing the class average to 96. Many of the answers shared the same distinctive ChatGPT-generated reasoning. When the final returned to an in-person format, the average collapsed to 48, 27 students dropped the course, and 22 of them had previously earned perfect scores. The story isn't simply about AI-assisted cheating. It's about the collapse of a model of assessment built on the assumption that submitted work reflects human thinking. If AI makes it effortless to bypass the work of judgment, universities are left evaluating polished answers without knowing whether any learning occurred.

The common thread isn't that AI is making people less intelligent. It's that it's making uncertainty easier to avoid, and judgment easier to outsource.

🤖 Why Does It Need a Face?

Last week, a rural New York school district approved the purchase of "Sally," a $57,590 humanoid AI teaching assistant. Sally is built by Realbotix, a company that describes itself as designing "AI-powered humanoid robots built for meaningful human interaction."

The obvious question isn't whether Sally works. It's why the AI needs a face in the first place.

A laptop running the same large language model (LLM) could answer questions, explain concepts, and tutor students. The humanoid body doesn't fundamentally change those capabilities. What it changes is the experience. It creates the appearance of a relationship.

That isn't accidental. Realbotix grew out of a corporate restructuring that also included the maker of RealDoll, and today it markets humanoid robots for companionship as well as education. In both cases, the product isn't simply AI. It's simulated presence.

Whether Sally succeeds as a teaching assistant almost misses the point. Schools aren't just buying artificial intelligence. They're buying embodiment. The idea that learning feels different when knowledge looks back at you.

🪪 The Card Is Gone. The Checks Remain.

This week, Prime Minister Andy Burnham scrapped the UK's proposed digital ID scheme and redirected its £1.8 billion budget toward cost-of-living relief. It was one of his first acts in office, and hardly a surprise. Burnham had criticized the proposal from the start, and the Institute for Government noted that the plan had been rushed, poorly explained, and politically weak.

The part worth noticing isn't that the digital ID card died. It is about what survived. Right-to-work verification remains mandatory, and the government still plans to extend those checks to gig economy workers. Britain's broader digital transformation agenda also continues.

That's becoming a familiar pattern. The visible symbol, the card in your wallet or the app on your phone, draws the political fight. But the underlying verification systems keep expanding. The object disappears. The infrastructure remains.

🌿 The Understory

In 1770, Wolfgang von Kempelen unveiled a machine to the Habsburg court in Vienna. A wooden figure seated at a cabinet, dressed in Ottoman robes, that played chess. It beat Benjamin Franklin. It beat Napoleon. For most of eighty-four years it toured Europe and America while writers filled pages arguing about what was happening inside. Edgar Allan Poe wrote an essay in 1836 working out the mechanism by observation alone.

There was a chess master folded into the cabinet, sliding along a rail as the doors were opened one by one.

What's striking isn't the deception. It's how much intellectual energy the machine absorbed. Serious people spent decades reasoning carefully about a question (is this thing thinking?) that had a human being sitting inside it the entire time. The cabinet was doing exactly what it was built to do. It made a person invisible and made a machine seem present.

We've inverted it. The Mechanical Turk hid a human to sell the appearance of a machine. Sally hides a machine to sell the appearance of a human. The AI in a take-home exam hides a person's absence rather than their presence. In every case the cabinet is the product, and the question of what's inside is the part we're invited not to ask.

Kempelen, by most accounts, found the whole thing tiresome. He'd built it to win an argument at court and spent the rest of his life being asked about it while he tried to work on speaking machines and steam turbines. He is said to have called it a bagatelle, a trifle. The audience wanted the cabinet. He wanted the mechanism.

See you next Wednesday. As always, my email is hello@wiobyrne.com.

Digitally Literate is a reader-supported publication. To receive new posts and support my work, consider becoming a free or paid subscriber.

🔗 Navigation

Previous: DL 439 • Archive: 📧 Newsletter

🕸️ Connected Concepts:

• Veil of Sincerity — The surface of a text or system signals that it can be trusted, and we evaluate the veil rather than what sits behind it; the signal's direction can invert while the habit of judging by appearance stays intact.
• Automation Bias — The documented tendency to defer to a system's output over one's own reasoning, producing higher confidence and lower accuracy at the same time.
• Epistemic Uncertainty — The willingness to suspend judgment and say "I don't know," and what happens institutionally when that admission becomes rare.
• Assessment Validity — Whether submitted work still reflects the thinking it was designed to measure, and what an institution does when it can no longer tell.
• Anthropomorphic Interface — Giving a system a face or body to create the appearance of relationship, independent of any change in what the system can actually do.
• Function Creep — Verification systems expanding past their original purpose after the visible object that drew political opposition has been removed.
• Infrastructural Invisibility — Technologies becoming harder to see as they become more consequential, and the loss of the vocabulary needed to contest them.