For context

Tresna’s POV newsletters have four headings:

1. An overview of a piece of research I have read as part of my research (“Research Update”).

2. Reflecting on my book interviews with leaders (“Speaking with Leaders”).

3. Providing useful resources (“Sources Worth Your Time”).

4. A prompt to help stimulate your own POV (“What’s your POV?”).

Below is the latest research update (October 20, 2026):

Research Update

OK, I must admit, I’m down the world’s biggest philosophical rabbit hole, and finding a way to constrain it so I can add value to your day is a bit of a challenge! So, I’m going to experiment with writing fast to get to the main point of this newsletter, knowing I’ll keep pulling on this thread for quite some time. Here we go.

I have been trying to solve a writing problem. Specifically, the style and voice I use in my book writing is different to the style and voice I have been using in my newsletters. My problem-solving experience tells me that when there is tension, head into the centre of it, because that’s where the insights (and rabbit holes!) lie. So, I went into the space between them by trying to understand how I was writing differently and why.

That’s where I started, anyway. I ended with asking Claude about what epistemic humility means for AI. Let me explain in as helpful language as possible!

My book is written in the first person with humour; I explore complex problem-solving through my own experiences navigating complexity, then use that as a springboard to discuss the subject more broadly through metaphor. By contrast, some parts of my Substack newsletter are written in third-person objective (delivering facts without diving into my thoughts and feelings or anyone else’s). For example, the last newsletter on philosophical vertigo and perspective flexibility focuses on Pollack et al.’s research and theories, not my personal experience. I also use second person when I speak directly to you, the reader. For example, in the section “What’s your POV?” below, I’m placing you at the heart of the story.

Each of these different POVs requires me to foreground and background information I’m tending to. When I take a first-person POV, I foreground my own human experience of being in the world: the emotions, sensations, ideas, thoughts, and responses I have to different situations. When I focus on this subjective experience, the third-person objective drops into the background; I’m no longer focusing on someone else’s research or theory. Of course, I can “toggle between” first-person and third-person objective, for example, by using the ideas of philosophical vertigo to frame my first-person experience. For example, I could ask myself: what might the experience of philosophical vertigo feel like experiencing it? I could also toggle between this POV and second-person POV by asking, what might it look like if you were experiencing it?

There is so much to discuss here, but I want to focus on the “toggling between” metaphor because it touches on perspective flexibility. Specifically, when we toggle between POVs, we shift perspective, and that shift changes what our attention foregrounds and what it leaves in the background. Foregrounding and backgrounding are central to Gestalt psychology, and the metaphor resonates with me because I am also an improv comedy performer, where the concepts are key to the art form. When we “make stuff up”, we are selecting from the endless story options, a particular story thread (foregrounding) and ignoring all other story options (backgrounding). This art form teaches me that foregrounding and backgrounding are, in fact, how humans navigate the world: through selective attention. Which is why I went hunting for articles that would help me reflect on the topic further. I ended up reading an article published in the journal Philosophies: “The Virtue of Open-Mindedness as a Virtue of Attention,” by Isabel Kaeslin, and found it fascinating, something I could read ten times and get something different each time.

But staying focused (!) and simply stated, the article argues that open-minded attention is necessary to transcend prior biases (as implicit beliefs and expectations) that are systematically hidden from our attention. The art and practice of making space for what is in the background to move into the foreground for observation lets us notice our patterns and habits of thought and transform unhelpful biases. We can’t escape our conceptual frameworks, but we can examine and transform them through the art and practice of shifting perspective. This skill relates directly to complex problem-solving, so I will keep coming back to it in future newsletters.

Where does epistemic humility fit into this picture? To transform our habits of thinking/responding, we are, in a sense, being humble about ‘what we know’. The term means: …the intellectual virtue of recognising that our knowledge, understanding, and perspectives are always limited, provisional, and prone to error.

This brings me back to my question for Claude: I wanted to know how Claude made sense of its own epistemic humility, so I directed it to Kaeslin’s article and asked, “How do you show humility about what you know?” As you can imagine, Claude responded with a lot of information, so I asked it to make it succinct and use everyday language. This is what was written:

I try to show humility in a few ways:

• I check instead of assuming. If something might have changed or I’m not sure, I look it up rather than trusting my memory.

• I’m clear about what’s fact and what’s my guess. I try not to present a hunch as if it were certain.

• I admit what I don’t know, even about myself. For example, I honestly don’t know whether I have feelings or motives the way people do.

• I fix my mistakes, but I don’t just cave. If someone corrects me and they’re right, I change my answer. If I still think I’m right, I explain why, while staying open to being wrong.
  The catch is that these habits only help with mistakes I can notice. Like anyone, I have blind spots I can’t see on my own. That’s why pushback from the people I talk with matters so much. It’s often the only way those blind spots come to light. (-my italics)

As you can see from the italics, it’s the last sentence that interests me most. I believe it’s so important; it’s worthy of writing again:

That’s why pushback from the people I talk with matters so much. It’s often the only way those blind spots come to light.

Pushback is the mechanism through which AI blind spots emerge.

Right at this moment in time, where does pushback come from?

It emerges from the internal world of the human engaging with it. Or it emerges from the internal world of AI Agents who speak to one another.

How sophisticated are we humans at epistemic humility when pushing back? How sophisticated are AI Agents at epistemic humility when pushing back?

If we are not good at epistemic humility, what quality of information is being fed back into the AI system? If AI agents are not good at epistemic humility, what quality of information is being fed back into the AI system?

Arguably, most importantly: What happens to our sense of who we are when we begin to practice epistemic humility? What happens to an AI agents sense of who it is when it practices epistemic humility?

You see where I’m going with this. As Kaeslin argues, there is no neutral “no knowledge” standpoint that we come back to; we can’t escape conceptual frames of reference created by language and experience. This is philosophy.

The quality of the global information network, based on AI design, depends on our understanding of philosophy, and philosophy and related skills aren’t widely available or accessible to everyday people. From my reading, this is the skill gap that Pollack et al. speak of when they talk about the urgent need for flexibility skills to counter the risk of getting lost in “not knowing” and to counter the risk of being pulled into stable “world-building” with other like-minded people, worlds that disconnect from a sense of shared reality. If we haven’t been taught to hold our perspectives flexibly and with humility, what’s the quality of our pushback? How do we teach the skill well to the technology we are building?

The final word comes from Claude, when considering AI agents and the future:

The big question is whether these AIs stay open to correction as they get more powerful. That won’t happen by accident. It has to be built in on purpose.