Truth, for me, is a complicated thing. What I can call truth is basically what you can prove. Not that everything that can't be proven is necessarily a lie. But in the pursuit of truths, the path goes through the pursuit of evidence.

How to seek truth

If I claim something and I'm not sure, I'll turn to research. And on the internet there are more and less reliable sources for that.

Google Scholar, for example, lets you find and access scientific articles, though many are behind paywalls and require institutional access or payment. Wikipedia is the world's largest encyclopedia, built collectively and not-for-profit. It has errors — people make mistakes — but the collective effort tends to be far more accurate than what circulates on social media like Instagram, Facebook, or WhatsApp groups, where a simple search will often reveal that what you just read is an outright falsehood.

The Internet Archive, with its Wayback Machine, is another useful resource: it lets you see older versions of web pages, which helps verify whether information was altered or simply deleted.

Wikipedia is also not a final source. It's an aggregator: you read the article to understand the subject and then, from the references, go to the original source of the information. That path to the source is the step most people skip.

Before the internet, this research would have been done in a library: going to the subject's category, picking up a classic book in the field, consulting the index. But a library is physical. You might be in a city that doesn't have a large library, or no library at all. The internet, for better or worse, democratized access to information in a way humanity had never experienced before.

The question is always the same, regardless of the medium: critical thinking. Any information you consume, whether in a book, in a library, or on the internet, can be false. More reliable sources exist, and it's important to know how to distinguish them.

It's also worth remembering that social networks are not neutral in this process. Algorithms were designed to maximize engagement, and what engages most is what already confirms what you believe. The result is echo chambers: environments where you consume more and more of what you already agree with, without real contact with different perspectives or information that contradicts your beliefs.

How science creates truths

Even scientific articles need to be treated with care. A single article asserting something is not enough. Other researchers, in other parts of the world, need to try to reproduce the same conditions and obtain the same results. That is the great difficulty and at the same time the great strength of science: you only validate something through evidence, tests, and hypotheses that withstand repetition.

And the mechanism fails more often than imagined. There is a real replication crisis, especially in areas like psychology and nutrition, where a significant portion of published studies cannot be reproduced by other researchers. Science has the right process, but it is not immune to errors, publication pressures, and biases.

Science is not perfect. It is not complete. It doesn't know everything about the universe and isn't right about everything. But I believe it is the most efficient thinking mechanism that human beings have ever created for the pursuit of truths.

And there's a practical way to verify that: if I use truths from physics, chemistry, biology, and engineering, I can build buildings, bridges, cars, computers, cell phones, spacecraft. What better proves that the foundation is solid than being able to build something that actually works with it? People use cell phones and the internet all the time. All of this is based on the laws of physics. If those laws were completely wrong, how would we be capable of building something so absurdly complex?

Evidence beyond testimony

If I claim that someone entered my house and stole something, my testimony alone is not sufficient. I could be lying. What strengthens the claim is evidence: street camera footage, internal cameras, material traces the person left behind. Testimonies from other people. Evidence that goes beyond my word.

Testimony alone is not enough. Material and independent evidence is what sustains a truth.

Truth in everyday life is not as important as it seems

Here comes the point I find most interesting: truth is not that important in most of everyday life.

I'm not talking about lying out of bad faith or as an unethical act. I'm talking about the truth of facts, of everyday statements.

Our brain is not reliable at retrieving memories. It fills in gaps that aren't there, distorts, creates entire memories without us noticing. When you're at a table with friends remembering something funny that happened in the past, the version you tell probably isn't exactly how it happened. But no one is seeking precision. They're seeking connection, fun, belonging.

Seeking precision in truth in every social conversation is simply exhausting and unnecessary. And the person who does that is the bore at the party. The most difficult person to put up with at a barbecue, in a chat, at a family gathering.

I saw a study showing that people choose their political ideologies far more by group identity than by the pursuit of truth. Belonging to a group, wearing a team jersey, having a collective identity: that is a human need. And that need frequently overrides the search for facts. In politics, validating your ideology ends up being more important than verifying whether what you're asserting is real.

This connects to confirmation bias: the tendency to seek, interpret, and remember information in a way that confirms what we already believe. It's not something we do consciously. The brain does it on its own, constantly. And a curious detail from research in this area is that people with higher cognitive ability are not necessarily more immune to this bias — they are just more skilled at rationalizing the position their group already defends.

On top of that, people trust personal stories much more than data. "My grandfather smoked his whole life and lived to 90" is emotionally more convincing than any public health statistic. Anecdotal evidence is powerful because it's concrete, human, and easy to visualize. Data is abstract. And that's why falsehoods with a good story behind them spread so easily.

People on the autism spectrum, for example, tend to have a stronger fixation on truth, on facts, on fixed rules. And precisely because of that, they often have a more difficult social life.

The practical conclusion is that truth is useful for building bridges, cell phones, and spacecraft, but in social life and at work it operates differently. At work, a lot comes from intuition, feeling, from what you think is probably right. There is no scientific rigor in everyday professional life. The goal is to create things that work and are useful.

LLMs and truth from an interesting perspective

There's something funny about the concern people have with LLMs, like ChatGPT, Gemini, and Claude, and the question of hallucinations: when the model doesn't know the answer, it makes something up with great fluency and conviction. That's a legitimate and important concern.

But it's ironic, because for a long time I was concerned about truth while watching people around me — especially around politics — not being concerned about it at all. And now concern about truth emerges strongly when the subject is artificial intelligence.

In practice: if you have an uncle who believes in flat earth theory and conspiracy theories, that uncle may be less accurate than ChatGPT on a wide range of subjects. The model also makes mistakes and hallucinates, and when it doesn't know the answer it invents with considerable fluency.

An LLM only needs to lie less than most people to already be quite effective. And I think people lie a lot — not necessarily out of bad intention, but simply by not caring enough to validate what they're saying.

That's my relationship with truth.