At TED in Vancouver, I found myself asking a question that,
not that long ago, would have felt philosophical, maybe even abstract. Now it feels immediate, and harder to avoid.
So, what does truth mean in today’s world, and does it still function
the way we think it does?
The question first came up at a TED Brain Date. A small room, about a dozen people, no stage, no slides, just a circle that filled in as people realized what the
topic was.
The group was a mix you only really get at TED. A journalist and NYU professor working in Ghana, focused on how students understand truth in a fragmented media environment. A PricewaterhouseCoopers consultant from India thinking about trust, incentives, and how organizations make decisions when the underlying data is uncertain. A former
Reuters reporter now working on credibility scoring systems. An AI founder who has spent decades training language models and recently launched a product he described, somewhat provocatively, as a
“lie detector for AI.”
advertisement
advertisement
We went around and talked about why we were there. What emerged quickly was that no one was approaching truth as a purely theoretical concept. The journalist
talked about students who no longer default to institutional authority. They don’t begin with the New York Times or the BBC. They begin with what shows up in their feeds, or what aligns
with their worldview, and they build from there. That shift changes not just what people believe, but how they decide what is worth believing.
The PwC consultant framed it more structurally.
Truth, at least in the way he had been trained, is supposed to be binary. Something is either true or it isn’t. But in practice, especially in a world shaped by media and AI systems, that
clarity doesn’t hold. What people encounter instead are multiple versions, each supported by different data, different narratives, and different incentives.
The AI founder pushed
in a different direction. After years of working with large-scale language systems, he said his company had stopped trying to define truth as an output. Instead, they surface evidence and assign a
confidence signal, leaving the final judgment to the user. They’ve moved, in his words, from truth to trust.
From there, the conversation turned toward how we define truth itself.
There was general agreement that objective truth still exists: scientific facts, empirical observations, things that can be measured and tested. But that only held for a moment, because as soon as you
move from the fact itself to what it means, or whether it matters, you are in a different category.
The group began to describe three kinds of truth: Objective truth, grounded in evidence.
Intersubjective truth, things like money or institutions that exist because we collectively agree they exist. And then a more personal category, something closer to lived experience, what feels true
to an individual even if it cannot be universally verified.
Once that framework was on the table, it became clear that AI operates primarily in the first category, and only partially in the
second. The third category, subjective truth, remains outside its reach.
That realization led to another question. The need to define a single, fixed truth may itself be cultural. In more
individualistic, Western frameworks, there is a strong pull toward resolution, toward a single answer. In other parts of the world, there is more tolerance for ambiguity, for multiple truths
coexisting, for ideas that evolve over time rather than locking into place.
Later that week, the conversation picked up again in a very different setting.
Dinner at noted Indian
Vancouver restaurant Vij’s. Ten tables, about 80 people, long conversations that moved between courses, the kind of room where you lean in a little closer to hear the person across from you, and
then realize the table next to you is having a version of the same conversation. The topic was framed simply: “Does Truth Still Matter?”
What was striking wasn’t
disagreement. It was how quickly everyone engaged. Truth clearly mattered to everyone in the room. That wasn’t in question. What wasn’t clear was what kind of truth we were talking
about.
At one table, the conversation turned to whether there is such a thing as universal truth, or whether everything we call truth is filtered through language and culture. Someone argued
that we may only ever approximate truth, that we get close but never fully capture it, and that language, while necessary, also distorts. Another person picked up on that and connected it to the idea
of the Tower of Babel, that once language fragments, shared understanding fragments with it.
At another table, the conversation was more practical. A former journalist talked about sourcing,
about how every piece of information has an origin and an agenda, and how the work is not just to report facts but to understand where they come from. “Everybody has an angle,” he said.
“The question is whether you know what it is.”
At a third table, the conversation drifted toward AI and systems. One participant described the current moment as a shift from
answers you could question to answers that arrive fully formed. Another said that the real issue is not whether AI is right or wrong, but whether people are still doing the work to question it. A
professor described what he’s seeing in the classroom. Students are still willing to argue, but they are less practiced at evaluating information. “They’re not losing
disagreement,” he said. “They’re losing discernment.”
The conversations were, in the best sense of the word, frothy. They moved quickly, overlapped, contradicted each
other, and then circled back. But there were moments when the room came into focus.
One of those moments came when a speaker said that proof is the oxygen of democracy. The line didn’t
need emphasis. It just settled into the room, because everyone understood the implication. If that’s true, then the question becomes whether we still agree on what counts as proof.
And
that’s where the tension showed up.
We don’t just disagree on conclusions. We disagree on evidence: what is valid, credible, admissible. Without that shared baseline, the rest
becomes harder to sustain. If we can’t agree on proof, we can’t agree on outcomes, and if we can’t agree on outcomes, the system starts to strain.
What connected the Brain
Date and the dinner was not a sense that truth has disappeared. No one argued that. The concern was something more structural. Truth is harder to define, harder to agree on, and harder to maintain as
a shared construct.
The complicating factor is that we are living in a moment that rewards speed and certainty. Media cycles are faster. Platforms prioritize engagement. AI systems are
designed to produce clear, confident answers. All of which pushes toward simplicity at the exact moment when the underlying reality is becoming more complex.
That gap is where the tension
sits.
Truth still matters. That much was clear.
What kind of truth are we talking about? That’s where the work begins.