The Four Layers of Information Reality
A framework for understanding how information becomes belief.
Information doesn’t arrive in your mind fully formed. It travels through layers before it reaches you, and each layer transforms it along the way.
Consider a headline you read this morning. Before becoming a headline, that information existed as raw data somewhere: numbers in a database, quotes from an interview, observations recorded by someone on the ground. Then it was processed, filtered, and organized into a form that could be analyzed. Then a writer shaped it into a story with a beginning, middle, and implication. Then you read it and, if it resonated, it merged with things you already believed to be true about the world.
Each of those stages has its own rules, its own gatekeepers who decide what passes through, its own distortions introduced along the way. Some of those distortions are intentional. Others are not. By the time information reaches the form you encounter it, the original raw material has been transformed multiple times by systems and people you’ll never see.
Understanding these layers changes how you evaluate what you read, watch, and eventually believe. You start to notice where information came from and what happened to it along the way. Over time, a sense develops for which claims deserve confidence and which deserve scrutiny.
Think of information as existing in four distinct strata. Each one transforms what came before it, grows less visible than the last, and exerts more influence over your perception of reality.
This framework is a teaching model, not a complete theory of epistemology. The layers don’t operate in a clean pipeline. They loop back on each other in ways that matter, and those complications are addressed throughout. The goal is usefulness, not theoretical perfection.
Layer One: Raw Data
Raw data is the unprocessed material of the information world: every click on a website, every purchase, every location ping from a smartphone, every heart rate reading from a fitness tracker, every search query typed at 2am.
The scale of data collection has grown dramatically. Industry estimates suggest about 5 billion people worldwide now use smartphones, and each device continuously records GPS coordinates, app usage patterns, biometric readings, transaction records, and metadata attached to messages and photos. By some estimates, global data creation exceeded 100 zettabytes in 2024. For perspective: a single zettabyte equals one trillion gigabytes.
This layer remains invisible to the people producing it. The data exists in server farms and corporate databases, accessible to the companies collecting it but rarely to the individuals generating it. When you scroll through Instagram, you see posts. The platform sees something different: time spent on each post measured in milliseconds, scroll velocity, tap patterns, pause duration, which posts you linger on, which you skip immediately. You experience content. The platform collects data about your experience of that content.
Raw data on its own has limited meaning. A single GPS coordinate tells you nothing useful, and a heart rate of 72 at 3:47pm is just a number. A record showing that someone clicked on an ad doesn’t reveal whether they were interested, bored, or simply scrolling too fast. The value of raw data emerges only when it gets processed into patterns, which is where the next layer begins.
One complication: even raw data isn’t neutral. Someone decided what to measure and what to ignore. The choice to track time-on-page but not emotional response, or to log purchases but not the conversations that preceded them, shapes what patterns can later be found. Data collection is itself an act of selection, influenced by the beliefs and priorities of whoever designed the system. The “raw” layer is already, in subtle ways, cooked.
Layer Two: Processed Information
Processing transforms raw data into something that looks like a fact. A credit score takes payment histories, debt ratios, account ages, and credit inquiries, then runs them through proprietary formulas to produce a three-digit number. A social media feed filters billions of posts through engagement algorithms to show you a particular sequence. A search engine takes the entire indexed web and returns ten results on the first page, ranked in a specific order.
Processed information feels neutral because it arrives looking clean and authoritative. The credit score appears as an objective measure of financial responsibility, the search results as the best answers to your question, the recommended video as a natural next step based on your interests. The format suggests certainty. The presentation implies that some neutral process simply surfaced what was already true.
The processing itself, however, involves choices at every step. When FICO calculates a credit score, it weights certain factors more heavily than others: according to FICO’s published methodology, payment history accounts for approximately 35% of the score, while new credit inquiries account for approximately 10%. When Google ranks search results, it prioritizes factors like backlinks from other sites, page loading speed, mobile optimization, and user engagement patterns. When TikTok builds your For You page, it optimizes for watch time and engagement, meaning content that keeps you scrolling rises while content that makes you close the app falls. These choices reflect the priorities of whoever designed the system, whether those priorities are predictive accuracy, advertising revenue, user retention, or something else entirely.
The critical feature of this layer: the processing happens out of sight. You see outputs, not the decisions that shaped them. FICO’s formula is proprietary. Google’s ranking algorithm is protected. TikTok’s recommendation logic isn’t disclosed. The processed information arrives with an aura of objectivity, but the processing itself remains hidden.
The boundary between this layer and the next one blurs more than the framework suggests. A graph showing unemployment trends over time is processed information, but the choice of timeframe, scale, and visual presentation already implies a story. A ranked list of search results is technically just sorted data, but the ranking itself shapes what narrative you’re likely to encounter. Processing and narrative-making often overlap.
Layer Three: Narrative
Narrative is information assembled into story: the news article explaining why the stock market dropped, the documentary tracing how a crisis unfolded, the text from a friend summarizing what happened at the meeting you missed.
This layer is where much of daily information consumption takes place. Reading raw data requires technical skills and access that can be difficult to obtain, and evaluating processed information requires visibility into systems that are often proprietary. Narratives package everything into something consumable. They provide characters, causation, sequence, meaning. A narrative takes the complexity of the world and makes it legible.
Narratives do something the previous layers cannot: they tell you what the information means. A narrative takes processed information (”unemployment rose 0.3%”) and gives it significance (”the economy is struggling” or “a healthy correction” or “the result of failed policies”). The same data point can anchor completely different stories depending on who tells them and why. A pharmaceutical company and a consumer advocacy group can look at the same clinical trial results and produce narratives that barely seem to describe the same study.
Every narrative involves selection, and selection means exclusion. A journalist writing about a scientific study chooses which findings to highlight and which to leave out. A documentary filmmaker sequences interviews and places emotional beats in ways that shape interpretation. Even an algorithm-generated news summary decides which sentences from a longer article deserve to represent the whole. These choices may be made carefully or carelessly, with good intentions or bad ones, but they are always made. The story is never the complete picture, and it can never be complete. What gets compressed out is often as significant as what remains.
Narratives also feed back into the earlier layers. A dominant narrative about what matters (say, that engagement is the best measure of content quality) shapes how processing algorithms get designed. A narrative about what’s worth studying shapes what data gets collected in the first place. The layers don’t flow in one direction; they influence each other continuously.
Layer Four: Belief
Belief is where narratives go to become invisible.
A belief is a narrative that has been absorbed so completely it no longer feels like something learned. It feels like something known, something obvious, something that doesn’t require justification because it simply is. The sky is blue. Hard work leads to success. Certain foods are healthy. Certain sources are reliable.
This layer holds the most power over perception and behavior, and it receives the least scrutiny. By the time information becomes belief, the path it traveled has usually been forgotten. The narrative that installed it has faded from memory. The processed information underneath has become irrelevant. The raw data is long gone. What remains is a sense of knowing that feels more like instinct than conclusion.
Some beliefs are well-founded. They traveled through rigorous layers: carefully collected data with transparent methodology, processing that was documented and reproducible, narratives that acknowledged uncertainty and presented competing interpretations. These beliefs earned their place through a chain of custody you could reconstruct if you tried.
Other beliefs arrived through thinner paths: data collected with built-in bias, processing designed to produce a predetermined outcome, narratives constructed to persuade rather than inform. These beliefs were installed by repetition or emotional resonance or social pressure rather than by evidence carefully weighed.
Both types feel equally true to the person holding them. Belief doesn’t come with provenance attached. The sense of certainty that accompanies a well-founded belief is indistinguishable, from the inside, from the sense of certainty that accompanies a belief with no foundation at all.
Beliefs are necessary for functioning in daily life. You can’t investigate everything from first principles every time you make a decision. But beliefs that go unexamined tend to accumulate, and the sources that installed them fade from view.
This layer is also different in kind from the first three. Data, processed information, and narrative are all types of external information. Belief is a psychological state. The framework treats them as continuous, but the shift from layer three to layer four crosses a boundary from the outside world into cognition. The skills for evaluating external information (sourcing, verification, contextual analysis) differ from the skills for examining internal beliefs (introspection, intellectual humility, willingness to update).
Why the Layers Matter Now
These four layers have always existed in some form. People have always collected observations (data), organized them into useful patterns (processing), told stories about what they meant (narrative), and internalized those stories into worldviews (belief). Similar frameworks have been articulated before, including the DIKW hierarchy (Data, Information, Knowledge, Wisdom) developed in library and information science in the 1980s.
What changed is scale and speed.
Data collection has expanded beyond previous limits.
In 2010, the average person generated measurable data through credit card transactions, medical records, and occasional computer use. In 2026, the average person generates continuous streams of behavioral, biometric, and location data through devices they carry everywhere. For example, the smartphone in your pocket records where you are, how fast you’re moving, what you’re looking at, and how long you look at it.
Processing has become automated and opaque.
The algorithms that determine what information reaches you, whether search results, social feeds, or news recommendations, operate at scales no human could replicate and through methods that companies treat as proprietary. You can’t examine the code that shapes your information environment. You can’t audit the weights. You can’t reliably test the system, because the outputs are personalized to each user.
Narrative production has multiplied.
In the early twentieth century, narrative sources were relatively concentrated: newspapers, radio broadcasts, national magazines. Today, anyone with an internet connection can construct and distribute stories to large audiences. The number of people producing narratives has grown from thousands to millions.
Belief formation has accelerated.
The time between encountering new information and integrating it into existing beliefs has shortened. A story can spread globally within hours. Repetition that once required years of cultural circulation can now happen in days through algorithmic amplification.
The machinery operating across all four layers is vast and interconnected.
Practical Application
Understanding the layers creates specific questions at each stage of information’s journey.
→ When encountering any claim, identify which layer you’re looking at.
Are you seeing raw data, like numbers from a study or records from a database? Processed information, like a credit score or a ranked list of search results? A narrative, like a news article or explainer video? Or is this something you already believe and are simply confirming? The questions appropriate for each layer differ.
→ When encountering processed information, ask who did the processing and what they were optimizing for.
A credit score optimizes for predicting default risk to lenders. A social feed optimizes for engagement and time on platform. A search engine optimizes for relevance as defined by its designers. The output reflects those priorities.
→ When encountering a narrative, ask what got left out.
Every story requires selection, and selection means exclusion. What data did the storyteller lack access to? What interpretations did they set aside? What would a narrator with different incentives have emphasized? What would change if the same information were presented by someone with the opposite conclusion?
→ When noticing a belief, ask how it got there.
Can you trace it back to a specific source? Did that source have access to good data, transparent processing, and a reason to tell you the truth? Or did the belief simply arrive one day and stay?
→ When any layer feels too clean, look for the feedback loops.
Is this data being collected because of an existing narrative about what matters? Is this narrative gaining traction because it confirms beliefs people already hold? Information rarely flows in one direction. Asking where the loops close can reveal assumptions that would otherwise stay hidden.
Total skepticism leads to paralysis. Total credulity leads to manipulation. This framework offers a middle path: know which layer you’re operating in and what questions fit that layer.
The Limits of Layers
This framework is a mental model, not a scientific taxonomy. The boundaries between layers blur in practice. Processing involves choices that could be called narrative. Narratives contain beliefs baked in from their creators. Data collection itself is shaped by beliefs about what’s worth measuring. The categories exist to make something complex easier to examine, not to carve reality at perfect joints.
The linearity of the framework is also a simplification. Presented as four strata, the model implies information flows downward from data to belief. In practice, the layers loop back on each other constantly. Beliefs determine what data gets collected. Narratives shape how algorithms are designed. Processed information influences which narratives gain traction. The pipeline is really a cycle, or more accurately, a tangle of cycles influencing each other simultaneously.
The fourth layer, belief, sits differently than the first three. Data, processed information, and narrative are external. Belief is internal. Grouping them together is useful for tracing how information moves from the world into your mind, but it glosses over the shift from information evaluation (a skill you can practice) to belief examination (a harder, more personal process that involves identity, emotion, and cognitive biases beyond the scope of any simple framework).
This framework also can’t tell you what to believe. It can help you see the path information traveled, but it won’t automatically reveal which paths are trustworthy. A well-sourced narrative from a rigorous outlet can still be wrong. A viral rumor with no clear origin can turn out to be true. The layers provide structure for asking questions, not a formula for arriving at answers.
What the framework offers is a way to slow down. When information arrives that seems important, the layers provide a structure for asking: Where did this come from? What happened to it along the way? And do I want to let it become part of what I believe?
The questions don’t get easier with practice, but they do become habit. And habits shape perception, which shapes belief, which shapes everything else.
Brilliant! Best ever thing I’ve ever seen on substack … for that matter, on this topic among the very best anywhere. Would that everyone studied these essays.
This is not a compliment. It’s an honest evaluation.
This essay was so important for me to read as a middle school librarian because I’m now thinking about student research and what layers of information my students can both access and analyze. Most of their research involves looking at narrative, so asking them to step into the earlier layers requires specific instruction on how to interpret data and adds a new level to source evaluation skills. Thank you for writing this!