The basic units of knowledge are not facts or observations

Instead, human knowledge is built on collections of stories, theories or principles

In a recent Substack, LM Saracas offers a short and thought-provoking take on today’s online information dynamics. For example, he argues that "we are all conspiracy theorists now" because we “are all in the position of holding beliefs, however sure we may be of them, that some non-trivial portion of the population considers not just mistaken but preposterous and paranoid."

His bases his analysis on the differences between how we experience information online and how we make sense of it. We encounter information online as a vast array of loosely connected but discrete data points (that he calls the 'Database') but, as humans, we make sense of the world via Narratives. Problems arise because the Database allows many compelling, but mutually incompatible, Narratives and provides no way of definitively choosing between them. Each of us therefore interprets what is happening via our own Narratives and we struggle to see or feel the others.

Importantly, this distinction between how we encounter information and how we make sense of it isn't new or restricted to the internet or social media. My recent post on the history of science showed the same distinction emerging within the practice of science.

Many of the original natural philosophers, particularly Francis Bacon and followers, built their work on careful observations. They tried to collect discrete pieces of information or data points and waited for the patterns of truth to emerge. Over time, scientific practice changed so that it concentrated on developing theories and testing these against data, observations and therefore the world. Like Narratives, Theories structure information in a way that makes sense of, and can explain, the world.

One of the reasons for the change was that people realised inductive generalisation from data points to truth was hard to justify and rarely definitive. Often, more than one theory or structure could fit the observations, just as multiple Narratives make sense in the Database.

In both situations, data and observations are not meaningful unless they are situated within a Narrative or Theory - within some kind of structure that allows us to make sense of or explain the world. This suggests that the foundational components of human knowledge are not the data points, observations or even facts, but abstracted structures like Narratives and Theories.¹

The foundations of knowledge

For some readers this might seem obvious, while others will find it highly counter-intuitive or just plain wrong. Surely, when done properly, knowledge, science and research is based on careful observations and built on top of known facts? And this means that facts and observations are the foundations of knowledge?

While we clearly should use observations, data and facts to justify or demonstrate something as knowledge, this does not mean they are foundational to the structure of knowledge. For the 'thing' we need to justify (via observations or data) is an abstracted structure, like a Narrative or a Theory. Let me offer a few observations to explain why.

1. As noted previously, in the language and practice of science, major new scientific discoveries take the forms of theories or abstract structures. To pick one example, human biology was revolutionised in the early 1950s by identifying the double-helix structure of DNA. This was not an observation or a set of new facts, but a description of a mathematical structure that described existing facts.

2. Scientific observations, as per Pierre Duhem, are almost invariably 'theory-laden'. That is, the methods or machines we use to make observations depend on existing scientific theory. Duhem used the example of measuring temperature and he described the way various physical theories underpin our observations. But this occurs across all fields of science. To pick a different example, if you are observing plants for a botanical study, your observations will focus on a series of characteristics that are important to botanical science, not just any random observations.

3. In addition, many of the terms or concepts we use to record or describe observations are also ‘theory-laden’. They only make sense in the context of some theory. What does a temperature of 295K mean if we don't have the underlying physics? Or a statement that GDP has grown by 2.1% in 2022 without the underpinning economics and statistics? The facts and observations we regularly rely on literally don’t make sense without a range of underpinning theories.

4. Learning new theories, knowledge or adopting a different narrative can change the way we perceive the world. We can see this most amusingly in the various visual illusions where the same picture can appear to us as two different things (perhaps a duck or a rabbit) depending on what we think we are looking at. LM Saracas offers a different, but related, example in an essay on attention, where he argues what we know can change what we pay attention to and therefore the way we experience the world. To pick one example of his, learning different bird songs can lead to noticing and hearing the world differently. Our accepted theories change what we observe, so observations cannot be the foundations of knowledge.

5. It is rarely noticed that humans often teach children theoretic explanations rather than facts as the foundations of knowledge. For example, we teach our children that the sky is blue and grass is green; even though both statements are only factually correct in certain, limited, situations. What abstract assertions like this do is provide the foundations of theories and narratives that help children understand and know the world around them; even if they are factually problematic.

So if human knowledge is not built from facts, information or observations, how does it work?

As humans, what we (think we) know is structured and articulated as a story, a theory, or some other structure like a set of principles. All of these take features of the world and arrange them into an intelligible structure. In this sense they are abstract - they paint a picture or build a model of part of the world rather than describe the world exactly as it is. The world is too complex, both big and small, for us to get our minds around it.

Our knowledge, or what we consider we know, is typically a collection of stories, theories and principles that we have adopted from a range of different sources. Some are embedded in our cultures or religions or absorbed from those around us. Others are taught explicitly through education. Some are our own attempts to make sense and explain aspects of the world. Some knowledge even seems to be hardwired into us - perhaps fire is dangerous or (to follow Immanuel Kant’s categories) that causality exists.

This structure also describes our body of scientific knowledge. This is a collection of theories, principles and (the occasional) stories that have been sourced and passed on by a wide range of sources. These are all abstractions from the world - most commonly theories or models - that describe some aspect of the world. We communicate and educate people about science by articulating scientific knowledge as theory or model, not as a collection of facts.

Understanding human knowledge as a collection of abstracted structures helps explain human belief formation and revision. For example, it is often commented that people don't simply change their mind, or accept that something they think is wrong, when provided with contradictory facts or evidence. One reason is, as we have seen, the facts or observations are not foundational to their knowledge. Instead, changing your mind involves changing the story, theory or principles that you use, not just to describe the world, but also to make sense of it.

Changing your mind is therefore personally costly and often requires you to see, hear or experience the world in a different way. It is therefore rational to be cautious about changing your views and knowledge. If you changed it every time different evidence came up, your knowledge would be in a permanent state of flux and you would struggle to function. But this sensible cautiousness very easily tips over into a stubborn refusal to admit that we could be wrong and the world is not as we think it is.

So, whenever you are trying to convince someone that their knowledge is wrong, it is worth trying to focus on how you can challenge or improve the stories or theories that make up their knowledge. Structure your points in a narrative or theory, and don’t just try to convince them of new facts or evidence. I can't guarantee you will succeed but expect at least you will have a more productive conversation.

And, for all of us, remember that what we think we know is structured as a collection of stories, theories or principles that partially describe the world. They are likely inaccurate in some way, and are definitely not a complete description. In other words, knowledge is hard and you are likely wrong in some way - although it is hard to know where. So, as always, it pays to maintain significant epistemic humility.

1. There are a range of different types of structures that would qualify here. We could equally talk about worldviews, philosophies, principles, proverbs, ideologies or pictures. We will focus on narratives and theories for simplicity of explanation.