Source of Authority (SoA)

Source of Authority in Narratives

Source of Authority is one of the key concepts of both AI Product Recommendation (AIPR) and Narr Theory.

Particularly for AIPR, narratives like “product X is best for ABC because…” is experienced and anecdote-based from first-hand experience, rather than completely objective like “1 + 1 = 2” or the concept of gravity.

To answer “how to become a better athlete”, I may look at what successful, professional athletes do, hear their stories, and listen to their advice.

In this example, the observed professional athlete becomes one of the Sources of Authority in the question “how to become a better athlete”

This is a very important topic to explore both how humans perceive ideas, create beliefs, and pass on narratives.

It is also a key component to understanding how AI recommends products.

---

Source of Authority (SoA) in Truth

How do we determine truth in real life?

This is a philosophical question of epistemology – “how we know what we know”.

There are different types of truths – rather, different natures of truth.

This is what we call Epistemology of Truth.

There are 3 epistemic establishment of truth.

• First-Party Established Truth
  • e.g. Sally’s favorite ice cream flavor – can only be formally “established” by Sally herself. Whatever Sally claims is her favorite flavor is the “truth” in this matter.
• Observed Truth
  • Empirical Truth (e.g. gravity)
  • Experience-based Truth (e.g. If I criticize David, he’ll likely get defensive)
  • Historical Truth
• Logical Truth
  • e.g. If the water is leaking because of the pipe damage, it won’t leak anymore if we repair the pipe

These 3 epistemic elements of truth, and combinations of these elements determine many expressions of truth we use today:

• Empirical truth (e.g. gravity)
  • Observed truth + logical truth (math, language, icons)
• Historical truth (observed truth with time)
• Mathematical truth (logic expressed in numbers, a universal language/icons)
• Expert endorsed truth (observed truth)
• Experienced based truth (observed truth)
  • e.g. If I criticize David, he’ll likely get defensive from my past experiences

In order for us to stay on course with AIPR as the main subject, we will keep the explanation on Epistemology of Truth simple here.

Narr Theory may post more articles on the Epistemology of Truth later.

For now, the main idea is that these are the different types of truth that humans can perceive at a high level.

Since humans are developing AI, the programming to which how AI finds truth will reflect how we perceive it.

Simultaneously, since AI thrives at pattern recognition, it may help provide more clues to help humans better understand the epistemology of truth.

Contradiction of Truth

This is worth noting in the context of AIPR.

Simply put, we can see when there are contradictions to truth.

Two sources may conflict or contradict the truth.

For instance, if Sally says she does not like mango ice cream. (First-Party Established Truth)

But she was seen eating mango ice cream 4 times this week.

This would contradict what Sally said herself about her favorite ice cream flavor. It would decrease our confidence in “knowing” that she indeed does not like mango ice cream. Our logic would indicate so.

Contradiction of Truth in AIPR

Let’s say that AI found narratives from its web search that Product X is the “most reliable”.

However, AI also finds narratives from others saying Product X “breaks down easily” – this would be a contradiction that could decrease AI’s confidence in establishing “Product X is a reliable” as truth.

---

Source of Authority (SoA) in AI Product Recommendation (AIPR)

Before AI recommends a product to a consumer, there is a very important step.

After understanding the user’s intent and their decision narrative, AI needs a process in which how AI will “determine truth”.

If AI is asked a question like “what product do you recommend for XYZ?” it will try to answer as best as possible.

However, this type of “product recommendation” truth is not absolute in nature.

• A more absolute truth will be something like (1 + 1 = 2) as example.

Even in the human world, a certain “recommendation” usually comes from lived experience, which then turn into stories, narratives, then passed onto another.

Since AI does not have first-hand experience like humans, it will scan the web and other digital content for evidence to strengthen its confidence in its recommendation.

To do this, AI finds sources of evidence to corroborate the truth.

If AI finds that many credible sources are supporting that “product X is the best for XYZ”, AI’s confidence in its recommendation would strengthen.

On the contrary, if AI detects a large pattern of contradictions with the claims a product makes, it would likely not recommend the product to a user. 

• E.g. Many products exaggerate certain features or even false claims like “instant setup”, “hassle free”, “extremely durable”, or “super quiet”
  • If this observation is not true from customer experience, there may be evidence online that contradict the manufacturer’s claim. AI knows this now and will continue getting better at it.

This is where it is important to understand Source of Authority – the sources that AI uses to corroborate the truth.

---

Types of Source of Authority (SoA) in AI Product Recommendation

AI tends to understand the context of product recommendation. If it is medical-related, it will proceed with caution, if it is more taste-based (e.g. fashion) it may look more at human narratives, if the product nature is more technical, it may lean more towards finding technical details.

At a higher level, thematically, here are the types of SoA’s in AIPR.

SoA as Owner of Truth

• A.k.a. First-Party Establisher of Truth

Tink of government official websites, company publishing, or just asking your friend what their favorite color is. Here, the truth is dependent on the first party’s definition.

• If AI was asked about a certain law in Toronto about sidewalks, City of Toronto website will have the exact “truth” to that answer. 
• If AI was asked about a certain company’s financial earnings or mission statement, the company’s website and publication would be the source of truth.

This type of truth cannot be refuted easily – simply based on the fact that the truth in question is coming from the source of truth itself.

Sometimes, however, if the truth is contradicted by the source itself, the truth loses consistency and therefore confidence.

E.g. Sally claims that she does not like vanilla ice cream. She does not enjoy the taste.

However, if Sally was seen having vanilla ice cream 4 times this week, there would be reason to say that the evidence contradicts her claim.

Same way – AI is capable of understanding the intent of manufacturers’ motif to claim that their product is the “best at XYZ” – AI is able to see this type of bias and account for it. Going forward, it is likely that AI may continue to improve on detecting intent of various parties in the human world in relation to one another.

———————————-

SoA as Trusted Institution

• A.k.a. “the association”

Medical associations, NASA, World Health Organization, Bank of Canada, U.S. Census Bureau, S&P Global, and even Consumer Reports – these are entities that were established for the purpose of educating the public on certain topics.

AI considers these as strong SoA due to their nature of official credibility, and the greater intent of public good.

———————————-

SOA as Perceived Expert

• A.k.a. “what the experts say”

This would be one of the stronger sources of authority both in real life and AIPR.

We often get ideas from athletes, professionals, doctors, leaders, and those that are considered to be experts in their field.

A toothpaste endorsed by a dentist, or a group of dentists would strengthen AI’s confidence in recommending the product.

However, in the context of AIPR, it is worth understanding “how does AI perceives an entity or an individual to be an expert in their field?”

This question can be both highly intuitive to answer and yet worthy of a deeper dive. Narr Theory discusses this topic of Perceived vs. Real Expertise here. (coming soon)

———————————-

SoA as First-Hand Experiencer

• A.k.a. “What others say“
• Customer Experience (CX), Voice of Customer (VoC)

When people buy a product, they go through the experience of using it for the intended purpose.

They may have been promised something in the product description, or have had certain expectations from what their friend said about the product.

Jason tells Sarah that “shoe brand ABC is comfortable for running”.

Sarah buys ABC. However, she experiences discomfort.

This could be due to Sarah’s orthotic difference from Jason, or perhaps they run different.

Nonetheless, Sarah experiences the product first-hand, which could either reinforce the product-narrative passed from Jason, or weaken or contradict it.

Where might Sarah go to share her experience? Potential destinations include:

• Other people
  • + Social circles
• Social media (private vs. public post or comment)
• Review (e-commerce site)
• Reddit
• Focused forum (e.g. running forum)

Specifically for AIPR, we should consider where such first-user narratives can be found digitally.

Social media, review sites, e-commerce sites, reddit, and forums – these are all respected SoA’s that AI uses to corroborate the truth.

Consumers tend to share videos, photos, and/or comment on social media.

For this reason, the creators of AI have the motif to develop AI’s capability to translate non-textual content into textual content.

• Increasingly, auto-generation of captions for videos are emerging as of July 2026.

Weighing of First-Party User SoA

Narr Theory theorizes this this SoA as First-Party Experiencer increases in authority when the volume scales.

A single highly credible expert in the field may have a heavier weighing of authority in establishing truth in a narrative than a day-to-day consumer.

• Olympic track-runner vs. casual jogger.

However, when there are 1,000 consumers’ first-hand narratives vs. 1 Olympic runner, let alone 10,000 or 1,000,000 first-hand narratives, the narrative landscape could be influenced more by the masses.

Simultaneously, it is worth noting that day-to-day consumer narratives can be influenced by other SoA’s as well.

A consumer’s own narrative may be influenced by what they heard from an expert, if they perceive the expert to “know better”, despite their personal experience not aligning 100% with such narrative.

This is what we call the Dynamics of Narratives which we will discuss in a future post.

The key takeaway for SoA as First-hand User is that the platforms on which these narratives are shared digitally are used as a potential source of truth corroboration by AI in the AIPR process.

————————– 

SoA as Edge pusher (Theory)

• A.k.a. “the edge of knowledge”, “the frontier of knowledge”

This SoA type is theoretical – so please keep that in mind.

Compared with other SoAs here – the Edge Pusher type has more evidence and examples to collect before formally establishing this as part of SoA theory overall.

The concept of Edge Pusher is this.

In a certain domain of knowledge, we often run into the edge.

• E.g. how did the universe come to existence?
  • Narrative 1: God created it
    • Question: Who created God?
      • Edge: We don’t know
  • Narrative 2: Big bang theory
    • Question: What happened before big bang theory?
      • Edge: We don’t know

Like this, humans reach different edges of knowledge in certain domains.

• E.g. (Semiconductors) – Silicon is the dominant semiconductor used in modern computing.
  • Is there a material that could ultimately outperform silicon in computing?
    • We don’t know.

The Edge Pusher isn’t necessarily the person who knows the answer.

They’re the person who expands the boundary of what humanity knows.

In life, these may be philosophical, religious, or scientific questions.

In day-to-day workflow, problems, solutions, and product discovery, these are long-tail narratives that a much smaller crowd in the audience may search.

Arguably, this small crowd may be influencers in their field’s narratives as the questions they ask ultimately become relevant.

Or, perhaps the long-tail questions are so seldomly explored that there is no commercial potential.

Theoretically, Narr Theory’s position on the Edge Pusher is actually shows signs of strong authority on the subject matter. Because one must know a lot about a topic (especially complex topics) to navigate to a reasonable, intelligent, and appropriate answer to a question that help push the boundary further to advance a certain purpose.

• “How can we better solve this problem?”
• “How can this product do X better?”
• “What sensible alternatives have we not tried yet?”
• “What is the underlying theme across it all?”

Edge Pusher in AIPR

When AI does not have an answer, it tends to do a web search.

Search Engine Optimization (SEO) systems, in a way, is an organization of knowledge (index). Narr Theory theorizes that both leading the frontier of knowledge and filling in gaps in established knowledge would be positive contributors to positioning a certain company or brand’s voice on the web in the subject matter.

Questions that are asked and searched frequently tend to have more saturated answers – both in SEO and AI retrieval.

Similar to how Amazon initially focused on helping consumers find long-tail books online that were not sold in-store, filling knowledge gaps and pushing the frontier may compound to establishing a certain entity, brand, company, or a product into a more credible position as a strong Source of Authority.

Amazon as of 2026 is perhaps… the largest e-commerce company.

---