Pathos Marketing
By Michael Hernandez | Where Psychology And Technology Converge

On The Power Of Patterns

Demystifying Algorithms And Large Language Models

By Michael Hernandez

Published: 3 January 2026

Introduction

Patterns unfamiliar
Patterns lead you through
To patterns of discovery tracing out the clues
Can you recognize the patterns that you find
Stuck in your mind?

-DEVO, "Patterns" from Oh No! It's DEVO (1982)

My favorite band, DEVO, sang of the power of patterns; How patterns dictate human behavior and how we can predict behavior if we pay enough attention.

And there is great truth in that observation.

Similarly, we can apply this perspective to the realm of algorithms and large language models (LLMs) which are intrinsically pattern-recognition and pattern-generation systems.

Algorithms, in the context of digital marketing, refer to the systems by which platforms surface content to users. These algorithms identify patterns in topical interests and engagement in users. These platforms serve content that mirror these preferences to promote continued user engagement.

LLMs, often conflated with the aspirational and more generalized label of "artificial intelligence," are designed to recognize the patterns of language and then generate constructs based on those patterns. There is no actual knowledge that these systems work with, rather the illusion of knowledge results from the statistical prowess of the LLMs' ability to predict what a response to a user's submitted prompt will be based on the content used in the materials it was trained on.

We could reflect from an anthropological perspective for a moment on why patterns are so important to humans: from the impacts of the seasons on agricultural projects to the reliability of ensnaring prey with baited traps, human survival was largely enhanced by the capability to recognize and leverage patterns in nature.

It makes sense that we would place great importance on patterns as we progressed into the era of the computer.

Philosophical musings aside, how can we leverage patterns in digital marketing? I think we can do so across two major dimensions: the technology, and most importantly the human.

Leveraging Patterns

Humans

Recognizing and influencing patterns of human behavior has been the core subject of academic, commercial, and political endeavor.

Academia influences patterns of thought to promote critical evaluation.

Commerce influences patterns of behavior to promote purchasing decisions.

Politics influences patterns of affiliation to promote votes and legislation.

Even religion influences patterns of belief to promote adherence and commitment.

If this topic interests you then I suggest the following resources for you:

Skeptical that humans are subject to the powers of patterns? Humans depend greatly on these cognitive shortcuts to reduce the burden of constantly thinking about everything.

Consider the following:

Jim is driving fast on the freeway. Faster than the posted speed limit. Driving 80 MPH on a stretch restricted to 65 MPH. When Jim recognizes that there is a California Highway Patrol vehicle directly behind him, he immediately reduces his speed to the posted limit.

We can observe this pattern of behavior with great consistency.

Too extreme?

Let's shift form an obvious reactive example to a subtler proactive scenario:

Jim is out strolling through a crowded park. For no reason at all, he pauses in the middle of the trail and proceeds to stare up past the tree line. Other guests in the park walking by stop as well and look up past the tree line. Not seeing anything they ask Jim what he is looking at. "Nothing." He responds, "Just wanted to see if anyone else would join me."

Don't think this would actually work? Have you ever tried prompting behavior like this before?

Well, I can tell you that from experience that is is kind of spooky to see it in action and work so consistently.

Because technology is a result of human action, it often reflects our dependence on patterns. Let us explore how we can leverage that foundation found in technology that provide cognitive shortcuts like the ones Jim demonstrated in our last examples.

Algorithms

Platform algorithms are designed to solve a human problem: retrieving relevant content from an overwhelmingly large library.

It is important to understand that fundamental situation.

Commercial incentives then influence bolt-on goals beyond solving that human problem: maximizing platform revenue.

The interplay between these two often lead to dark emergent patterns: the proliferation of negative news stories and inflammatory user-generated content. Users engage with polarizing content because it induces an emotional response.

That is not the route we want to pursue with ethical digital marketing projects.

We can often look to top-surfaced content on platforms to reverse-engineer what elements in content lead to promoted visibility by algorithms.

Here are the top areas I evaluate:

Identify the correlations. Revise your content. Deploy and measure.

If you see increased visibility, engagement, and interaction, then you have successfully reverse-engineered the algorithm to identify the elements they promote and reward.

You can now generate content that gains visibility more effectively knowing what elements you need to incorporate.

You can now generate content that will generate engagement more efficiently knowing what users respond to.

You can stop wasting time and effort on needless volume and focus on targeted quality.

Large Language Models

Large language models appear to demonstrate vast knowledge and reasoning capabilities. We see these tools act as copy editors, mathematicians, copy writers, visual artists, computer programmers, psychological therapists, and systems architects.

In reality, these are emergent from the core power to recognize patterns and predict the next item in a sequence, much like a glorified autosuggest tool.

Without diving too deeply into the complex computations and techno-jargon, here is a simplified flow of how LLMs work at a high level:

User Prompt ==> LLM
LLM [ Evaluates Prompt -> References Training Material -> Predicts Responses -> Evaluates Predicted Responses ]
LLM ==> Final Response

A user submits a prompt to the LLM. The LLM evaluates the prompt, identifying the words and the topics behind them. It then references the materials it was trained on to create predicted responses based on statistical relationships in how those topics are handled by those training materials. It then evaluates the fit of the predicted response and reiterates if needed. The final response generated by the LLM is a statistical marvel of patterns.

Understanding this process is fundamental in making productive use of LLMs:

Conclusion

Understanding how technologies work is fundamental in making effective and efficient use of them. Even as capabilities expand, such as introducing live web content access to LLMs, the basic framework can guide our thinking towards value extraction instead of passive distraction or manipulation.

Remember that these technologies were invented by humans to solve human problems.

Remember that these technologies often reflect our own pattern-based approach to thinking.

Remember that at there core these technologies are designed to surface content from overwhelmingly large libraries.

Remember that commercial incentives introduce bolt-on factors to promote platform engagement and retention.

Remember the patterns.

I close this article with the same words that DEVO sang to close their song with:

Old patterns never die; They just go on and on.

Thank you,

-Michael

Got Questions?

I might have answers! Check out the Ask Mike feature for answers to past questions or find out how to submit your own digital marketing question!