Simple cover by me

Probabilistic Approaches in Our Daily Life and Routine

When I decided to write about probabilistic thinking in my life, I read about machine learning and came across two main approaches: (I) Frequentist and (II) Bayesian. These methods are used to estimate and make predictions based on current data and context.

Before we begin our journey, it’s better that I give you some context. In physics, we have two major frameworks: quantum mechanics and Newtonian mechanics. Each operates under different rules, estimations, and formulas. So, when we want to make a guess or find an answer, or even make a decision, we must first understand the framework and the rules we’re operating within — in other words, our approach.

Now let’s move between a problematic and mathematical space. Imagine we’re working at the office, and suddenly a clown walks in to deliver our pizza order. What reaction would we have to this scene? In 99 percent of cases, we’d be shocked — not because of the pizza, but because of how weird it is to see a clown in that context. Why would we be shocked? It’s because our brain traces expected pathways and possibilities, and when the result doesn’t match those expectations, we’re surprised. Let me explain how the brain thinks — how it builds mental diagrams and makes predictions based on prior data. Over billions of years, our brain has been trained to respond by first searching for the most relevant data related to a situation. For example, when we see a clown delivering pizza at the office, our brain initially has no prior scenario to compare it with — we’ve likely never encountered anything like that. However, the brain must quickly find the most similar data to make sense of the event and assign a label to it — such as “interesting,” “scary,” “funny,” or something else. Once the brain categorizes the situation and finds the closest match, it decides on an appropriate reaction. If the brain can’t produce an appropriate reaction or assign a proper category, we tend to freeze — doing nothing, much like a loading page waiting for input.

deep thinking alert : Now, let’s talk about a child around the age of two — the stage when they begin to develop consciousness, think, and learn. While this process can technically begin from day one of life, for the sake of discussion, let’s assume it starts at age two.What approach this child will be couese ? He/she will suprised by clown or not ?

Now, let us analyze how a human — particularly a child — reacts in strange or frightening situations. This is important because modern machine learning (ML) and deep learning (DL) systems are based on similar principles: transforming data and structure from one form to another through experience.

When a child sees a clown for the first time, they may not react negatively — because, at that stage, everything is unfamiliar and appears “normal” in the absence of prior classification. In early development, a child continuously encounters various stimuli — such as other children, adults, trees, flowers, and common objects — and begins to categorize these inputs as part of a “normal” world.

During this phase, the child is essentially building a baseline or reference model of what is typical. After seeing hundreds or even thousands of human faces, the child starts to form a strong internal representation of what a “normal” face looks like.

However, when the child later encounters a clown — with exaggerated features, unusual colors, and a distorted facial structure — the experience conflicts with their established internal model. The clown’s appearance violates the learned pattern of normal human faces. This cognitive dissonance or pattern mismatch may trigger fear, confusion, or even distress. As a result, the child may start crying or show signs of fear.

We’ve read about two probabilistic approaches: (I) Frequentist and (II) Bayesian. Now, let’s decide which one we should use when making decisions for our child. How can we describe and demonstrate this situation?

Time to analyse , Bayesian or Frequentist ?

Senario 1: We can approach this situation in two ways. First, we can assume that a child uses a Frequentist approach — dividing all objects into two categories, like flipping a coin: heads or tails. In this case, things are either “normal” or “abnormal.” After seeing 2,000 faces (or even more), the child builds a probabilistic space and stores data in their mind. Using this accumulated data, the child can make guesses or respond to unfamiliar or scary situations. And when a child sees a clown with a scary face — based on the 2,000 faces they’ve previously seen — it doesn’t fit the “normal” category, so the child starts to cry.

Senario 2 :Now we come to the core idea and main goal of this discussion: the human brain is far more intelligent than we often assume. To survive in any environment, the brain must constantly protect the organism — not only from external threats, but also from internal confusion, sensory overload, and emotional disintegration. In other words, the brain is always defending itself, interpreting the world, and constructing adaptive responses to ensure survival. Even in the case of a child — who has not yet developed full cognitive abilities, self-awareness, or a stable sense of consciousness — we must recognize the presence of a powerful underlying system. Although the child may lack mature cognition, we cannot exclude them from analysis entirely.

We must postulate that any living entity with a brain possesses, at a minimum, memory, perceptual systems, and a capacity for internal modeling of the environment. Thus, every brain — regardless of age — operates not only by reacting to the world, but by actively constructing meaning, defending stability, and fighting for coherence and life. This process is not learned — it is inherent to biological intelligence.

A child — like the human brain in general — has memory. For example, if we observe a lion trying to survive against an animal stronger than itself, our minds process this as part of a broader “animality” dataset. Humans can transform and reinterpret unrelated data, making it relevant through analogy and pattern recognition. So, whenever we make a decision, we rely on prior data, and we also consider the current situation (likelihood) to find an answer — the posterior. This shift reflects a move from older probabilistic thinking toward more reliable data-driven models and postulation.

This is a Frequentist

As shown in the image above, we need to encounter a large amount of data in order to understand and learn patterns. The human brain is remarkably powerful in identifying patterns. However, stories often need to be more complete than the usual context for the brain to fully grasp them. In situations where no prior evidence exists, we can postulate that a child is capable of transforming unfamiliar data into meaningful associations.

In Bayesian estimation, before we make a guess about something, we rely on prior data to inform our expectations. For example, when we flip a coin, we know from past experience that the result will likely be close to a 50/50 split. This estimation is based on the assumption that the object has two sides of equal probability — reasoning supported by the data we’ve collected over time.

As shown in the diagram above, whenever we make a decision, we rely on three key components: the posterior, the prior, and the likelihood. These elements help us reason through uncertainty and guide us toward better, safer, and more reliable answers — whether we’re making a guess or a fully informed choice.

Machine Leaning

In machine learning, we often rely on data created by humans and users, and in simple scenarios, we accept this data as references for making predictions (as in basic machine learning models). However, sometimes we need to transform the data first, especially when trying to estimate or find an answer without clearly related data. In such cases, we can’t simply say, “Okay dear user, go ahead and drive at 300 km/h — and if you get into an accident, maybe choose 200 or 190 next time.” Instead, we need to create a structured space to make informed decisions — similar to how a child transforms raw input based on context and conditions to make a guess.

Farewell

The human brain, even at an early developmental stage like in a 2-year-old child, behaves in a probabilistic and adaptive manner — similar to Frequentist and Bayesian reasoning used in ML. As children gain experience, they form internal models of the world, and when something violates those models (e.g., a clown with an abnormal face), the brain reacts with surprise or fear. This suggests that the brain constantly builds and updates internal representations based on prior experience, likelihood, and context — just as Bayesian models do. Therefore, human cognition is not only biologically intelligent but inherently data-driven and survival-oriented.