✦ The Insight System

Truth in AI

How our AI actually works, what it can and can’t do, where bias hides, and how to be a smart consumer of AI-generated personality content.

← Back to The Insight System

How Your Report Gets Written

When you play The Insight Game and generate a personality report, here’s exactly what happens — step by step, no black boxes:

The Report Generation Pipeline
🃏
You Play
36 card choices determine your 4-letter type algorithmically. No AI involved here.
📋
We Assemble
Your type, the RoBards source material for that type, and your optional life context are combined into a structured prompt.
🧠
AI Writes
The prompt is sent to Meta Llama 3.3 70B running on Cloudflare’s private servers. The AI generates your report within the framework we provide.
📝
You Read
The report is returned to your browser and saved to your library. Your data is never stored by the AI or used for training.

What’s in the Prompt?

The AI doesn’t just get your personality type and guess. Here’s what we actually send it:

The “Recipe” — What the AI Receives

1. The RoBards Source Material: Core personality descriptions, temperament frameworks, stress patterns, work strengths, relationship dynamics, and growth edges — all written by Dr. Martine J. RoBards from 30 years of clinical practice. This is the foundation the AI builds on.

2. Your Personality Type: Your 4-letter code and which temperament family you belong to (Empathist, Analyst, Legalist, or Realist).

3. Your Context (Optional): If you shared your age, gender, occupation, relationship status, or satisfaction levels, those are included so the AI can personalize to your actual life.

4. Writing Instructions: Detailed guidance on tone (warm, direct, second-person), structure (specific section headers), length, and what to avoid (clichés, filler, repetition).

The AI does not browse the internet, access external databases, or know anything about you beyond what’s in that prompt. It generates the report from the RoBards framework applied to your specific type and context.

What Determines Your Type?

Your personality type is NOT determined by AI. It’s determined by a scoring algorithm that counts your card selections across four dimensions: Extraversion/Introversion, Sensing/iNtuition, Thinking/Feeling, and Judging/Perceiving. This is a mathematical process — the same answer every time for the same selections. The AI only enters the picture after your type is already determined.

Anatomy of a Prompt: What the AI Actually Sees

In AI, a “prompt” is the set of instructions and information you give the model before it writes anything. Think of it like a brief to a ghostwriter — the more specific and grounded the brief, the better the writing. Here’s exactly what goes into yours.

Every report prompt has four layers. Understanding them helps you evaluate what you’re reading:

📚
Layer 1: The RoBards Source Material
The Foundation — Written by a Human

This is the most important layer. For each of the 16 personality types, we provide the AI with Dr. RoBards’ original clinical descriptions: core personality traits, temperament classification (Empathist, Analyst, Legalist, or Realist), work strengths and challenges, relationship patterns, stress responses, and growth edges.

Why it matters: This is what separates your report from a generic ChatGPT response about “ENFPs.” The AI isn’t pulling from internet forums or Wikipedia — it’s working from a specific clinical framework developed over 30 years of practice.

// Example of what the AI receives:
Type: ENFP — The EMPATHIST
Core: “Endlessly inventive, imaginative, individually driven.
  Lives in the realm of the possible. Great starter who
  sometimes struggles to finish.”
Temperament: “NFs have a deep focus on relationships,
  meaning, and authentic human connection. Romance
  personified, endlessly seeking profound union with
  people and purpose.”
🧩
Layer 2: Your Personality Type
Determined Algorithmically — No AI Involved

Your 4-letter type (e.g., ENFP, ISTJ) is determined by a scoring algorithm that tallies your 36 card selections across four dimensions. This is pure math — the same selections always produce the same type. The AI never influences your type assignment.

The type then determines which RoBards source material gets loaded into the prompt. An ENFP report draws from ENFP-specific descriptions; an ISTJ report draws from completely different content.

👤
Layer 3: Your Personal Context
Optional — Provided by You

If you shared your age, gender, occupation, relationship status, partner’s type, or satisfaction levels, those are included in the prompt. This is what makes your report feel personal rather than generic.

How it’s used: The AI receives specific guidance like “Millennial: Emphasize early career, dating, identity. Acknowledge economic pressures” or “Occupation: Psychotherapist. Tailor work section to this specific role.” These are guardrails, not magic — they steer the AI toward relevant content.

What happens without it: If you skip the optional fields, the report still generates from the RoBards source material for your type. It’s just less personalized.

✍️
Layer 4: Writing Instructions
Quality Control — How to Write, Not What to Write

This layer tells the AI how to write, not what to say. It includes instructions like:

  • Write in warm, direct second-person (“you”) — like a counselor who knows the reader
  • Use concrete behavioral examples, not abstract generalizations
  • Never repeat the same idea in different words
  • Avoid filler phrases (“it’s important to note,” “at the end of the day”)
  • Use the RoBards temperament names: Empathist, Analyst, Legalist, Realist
  • End with the type caveat: “Your type describes preferences, not destiny”

Why this matters: Without quality instructions, AI defaults to generic, repetitive, hedge-everything corporate writing. These instructions are what make the difference between a report that feels alive and one that feels like it was written by a committee.

Putting It All Together

When you tap “Generate Report,” all four layers are combined into a single prompt and sent to the AI model. The model generates your report within those constraints. It cannot access the internet, look up additional information, or remember anything from other users’ reports. Each generation starts fresh from the prompt alone.

The Key Insight

The quality of an AI-generated report is almost entirely determined by the quality of the prompt — not the intelligence of the model. A brilliant AI with a vague prompt produces garbage. A capable AI with a rich, grounded, specific prompt produces something genuinely useful. That’s why the RoBards source material matters so much: it’s the difference between “tell me about ENFPs” and “here is 30 years of clinical observation about ENFPs — now write this person’s profile.”

How This Applies Beyond Our Platform

Understanding prompt structure makes you a smarter consumer of AI everywhere — not just here. When you use ChatGPT, Gemini, or any AI tool, ask yourself:

Where Your Data Goes (And Doesn’t Go)

✘ Most AI Apps
📝 You type something personal
→ Sent to OpenAI / Anthropic / Google
🏢 Processed on their servers
📄 Subject to their retention policies
❓ May be used for model improvement
✔ The Insight System
📝 You type something personal
→ Stays on Cloudflare infrastructure
🔒 Processed by private AI (Llama 3.3 70B)
🚫 Never used for training
✔ You own your data. Period.
What is Cloudflare?

Cloudflare is one of the world’s largest internet infrastructure companies, powering over 20% of all websites. Major banks, hospitals, and governments trust Cloudflare to keep their data secure. Our AI runs on the same infrastructure. Read Cloudflare’s official data usage policy →

Bias in AI-Generated Content

Every AI model carries biases from its training data. We take this seriously and want you to understand where bias can appear in your reports, so you can read them with informed eyes.

🌍 Cultural Bias

AI training data skews heavily toward Western, English-speaking, individualistic cultures. Personality descriptions may emphasize independence and self-expression in ways that don’t resonate with collectivist cultural values. If your cultural background is non-Western, some descriptions may feel off — that’s a limitation of the model, not of you.

⚖️ Gender Bias

AI models can reflect gendered assumptions about personality. A “Thinking” woman may get softer language than a “Thinking” man for the same type. A “Feeling” man may get more qualifying language. We actively prompt against this, but it can still surface.

💼 Socioeconomic Bias

Career advice in reports may default to white-collar, college-educated assumptions. If you’re in trades, service work, entrepreneurship, or non-traditional career paths, the AI may not fully reflect your reality. We’re actively expanding our prompt frameworks to address this.

🧠 Neurodiversity Bias

Personality typing was developed primarily for neurotypical populations. If you are neurodivergent (ADHD, autism spectrum, etc.), type descriptions may not capture how your personality manifests. Your experience is valid even if the framework doesn’t fully account for it.

What We Do About It

Our prompts include explicit instructions to avoid gendered assumptions, affirm non-traditional identities, and acknowledge cultural context. But prompts are not a complete fix for training data bias. Transparency about these limitations is more honest than pretending they don’t exist.

What AI Can and Can’t Do

✔ AI Can
Synthesize personality research into personalized language
Apply frameworks consistently across all 16 types
Adapt content to your demographic context
Generate novel combinations of established insights
Reflect patterns back to you from your own words
✘ AI Cannot
Truly understand your lived experience
Replace a therapist, counselor, or trusted friend
Account for trauma, mental health conditions, or medications
Guarantee accuracy for any individual
Know when it’s wrong (AI doesn’t doubt itself)

How to Read AI-Generated Personality Content

These guidelines apply to our reports — and to any AI-generated content about your personality, from any platform.

Green Flags — Signs of Responsible AI Use

Red Flags — When to Be Skeptical

The Human Foundation

AI is the delivery mechanism. The substance comes from a human being.

Dr. Martine J. RoBards spent over 30 years as a clinical psychologist studying how personality type shows up in real life — not in laboratory conditions, but in marriages, workplaces, parenting, and personal crises. Her Type-Temperament framework maps 16 personality types across four temperament families:

The Empathist (NF)

ENFP, INFP, ENFJ, INFJ — Driven by relationships, meaning, and authentic connection.

The Analyst (NT)

ENTP, INTP, ENTJ, INTJ — Driven by intelligence, competence, and mastery.

The Legalist (SJ)

ESTJ, ISTJ, ESFJ, ISFJ — Driven by duty, responsibility, and reliable contribution.

The Realist (SP)

ESTP, ISTP, ESFP, ISFP — Driven by freedom, action, and full presence in the moment.

When the AI generates your report, it doesn’t invent personality theory. It applies Dr. RoBards’ published descriptions, clinical observations, stress patterns, relationship dynamics, and growth frameworks to your specific type and life context. The AI is the pen. The wisdom is hers.

Your Rights as a User

You Always Have the Right To:

• Know exactly what AI model processes your data (Meta Llama 3.3 70B on Cloudflare Workers AI)

• Know that your content is never used to train AI models

• Request deletion of all your data at any time

• Disagree with your report and consider it wrong

• Use these insights as one input among many, not as gospel

• Contact us with questions, concerns, or feedback: michaelarobards@gmail.com

Our Commitment

We built this platform because we believe self-understanding matters and that AI can make it more accessible. But accessible doesn’t mean reckless. We commit to:

This page was last updated March 2026. If our AI model, infrastructure, or practices change, we will update this document and notify active users.

Read our formal AI Transparency Disclosure → | Privacy Policy →