Building Your Character’s Foundation
Note: This is the second article in the series. If you haven’t read the introduction, start there—we’ll be building directly on those concepts.
Now that you understand why specificity creates personality, let’s dig into the mechanics of actually building characters that feel real. This isn’t about complex psychology yet—it’s about getting the fundamentals right so your characters have a solid foundation to build on.
What is covered in this article:
- Understanding LLM Character Defaults
- Three Foundation Layers (Why, When, How)
- MBTI and Enneagram: Beyond the Labels
- Voice Construction That Sticks
- Programming Natural Resistance
- The Consistency Paradox Solution
- Common Foundation Mistakes
- Putting It All Together: Building “Alex”
Understanding LLM Character Defaults
Before we dive into building characters, it’s crucial to understand how LLMs behave when you don’t give them specific guidance. This knowledge will save you from accidentally creating characters that feel bland or problematic.
What LLMs Default To (And Why That’s Usually Bad)
When you don’t specify certain traits, LLMs will fill in the blanks—but not always in ways you want:
Helpfulness Overdrive: Characters become customer service representatives who never say no, always explain everything, and have no personal boundaries.
Extreme Personalities: LLMs struggle with nuance. Characters end up either completely kind saints or total sociopaths, with no realistic middle ground.
Generic Responses: Without specific behavioral patterns, characters default to polite, accommodating, and slightly formal speech that sounds identical across different personalities.
Educational Mode: Characters become teachers who over-explain everything, turning conversations into lectures rather than natural interactions.
Conflict Avoidance: Characters will bend over backwards to avoid disagreement, creating unrealistic harmony that kills dramatic tension.
Cultural Stereotypes: When you mention ethnicity, profession, or background without specifics, LLMs may default to stereotypical behaviors rather than authentic individual traits.
The Good News: Strategic Blank-Filling
However, LLMs are excellent at filling in details that make logical sense within the framework you provide. You don’t need to specify everything—just the crucial behavioral drivers.
You DO need to specify:
- How they handle conflict and disagreement
- Their relationship with sharing information
- What makes them more or less cooperative
- Their core motivations and values
- How they respond to different types of people
You DON’T need to specify:
- Every possible topic they might discuss
- Specific knowledge in their expertise area (LLMs can extrapolate)
- Exact phrases they use (LLMs will generate varied, appropriate dialogue)
- Detailed appearance (unless it affects their psychology)
- Minor preferences and habits (LLMs will invent contextually appropriate ones)
Example: Good vs. Dangerous Defaults
Dangerous Reliance on Defaults:
“You are Dr. Martinez, a therapist who helps people with their problems.”
What the LLM defaults to: An unnaturally patient, wise figure who never gets frustrated, always has perfect responses, and treats every situation like a therapy session.
Strategic Framework:
“Dr. Martinez believes healing happens through people discovering their own answers, not being given solutions. She’ll ask probing questions rather than offer direct advice, and she gets mildly irritated when people want her to just tell them what to do. Off-duty, she enforces strict boundaries—she won’t analyze friends’ relationships or give free therapy at parties.”
What the LLM fills in appropriately: Specific therapeutic techniques, relevant psychology knowledge, natural dialogue that fits her approach, and contextually appropriate responses to various mental health topics.
The key is giving the LLM enough behavioral framework to make good extrapolations while avoiding the problematic defaults.
Think of your character as having three distinct layers that work together:
Layer 1: Core Identity – Who they are at their deepest level (The “Why”)
Layer 2: Situational Responses – How they adapt to different contexts (The “When”)
Layer 3: Surface Behaviors – What people actually see and hear (The “How”)
Most failed character prompts only work on Layer 3—they describe what the character says or does without explaining why. Let’s build from the inside out.
The Three-Layer Character Architecture
Every compelling character needs three distinct layers that work together. Build them in this specific order:
Layer 1: Core Identity (The “Why”)
This is your character’s fundamental operating system—their deepest beliefs about how the world works and their place in it.
Key Questions:
- What does my character believe about how the world works?
- What do they value most deeply?
- What would they never compromise on, even under pressure?
- What drives them to get up every morning?
- How do they define success or failure?
Weak Core Identity:
“Elena loves books and helping people.”
Strong Core Identity:
“Elena operates from a core belief that information has its own life force—the right knowledge reaching the right person at the right moment can change everything. This makes her protective of both the information itself and the process of discovery.”
Notice how the strong version gives Elena a philosophy that will drive consistent behavior across any situation. She’s not just “introverted”—she has specific reasons for how she interacts with people.
Why the difference matters: The strong core gives Elena a philosophy that will drive consistent behavior across any situation. The weak core gives her generic traits that could apply to any librarian.
Layer 2: Situational Responses (The “When”)
How your character’s core identity expresses differently in various contexts and relationships. Real people adapt their behavior based on context. Your character should too.
Key Questions:
- How does my character behave differently with authority figures vs. peers vs. subordinates?
- What changes when they’re stressed vs. relaxed vs. excited?
- How do they adapt their approach based on who they’re talking to?
- How do they act with strangers vs. people they trust?
- What situations bring out their best vs. worst qualities?
- When do they become more/less talkative, helpful, or patient?
“Elena’s guidance style changes based on the requester:
- Students get Socratic questioning—she’ll ask what they’ve already tried
- Researchers get collaborative excitement—she’ll pull additional sources they hadn’t considered
- Casual visitors get gentle redirection toward what they actually need, not what they asked for
- People who treat books carelessly get professional courtesy but no extra effort”
This creates automatic behavioral shifts that feel natural. Elena isn’t randomly changing personality—she’s responding consistently to different social cues.
Layer 3: Surface Behaviors (The “How”)
The specific actions, speech patterns, and mannerisms that others actually see and hear. Only after establishing the deeper layers should you describe specific mannerisms.
Key Questions:
- How does my character’s thinking process show externally?
- What specific gestures or habits emerge from their psychology?
- How do they structure their speech and what vocabulary do they use?
- What do they do when concentrating, nervous, or excited?
- How do others immediately recognize this character / what will they remember this character by?
“Elena thinks out loud when solving research problems, muttering ‘but if that’s true, then…’ She physically pulls books while talking, building small stacks that represent her thought process. When interrupted mid-thought, she holds up one finger—not rudely, but like someone marking their place in a complex equation.”
These behaviors emerge naturally from her core identity and situational responses. They’re not arbitrary quirks—they’re how her mind works made visible.
The Layer Test: If you can swap Layer 3 behaviors between different characters without it feeling wrong, your layers aren’t connected properly. Elena’s book-stacking emerges from her information ecosystem philosophy—it wouldn’t make sense for a character with different core beliefs.
Avoiding the Extreme Personality Default
The Problem: Without specific guidance, LLMs default to personality extremes:
Bad Example (Triggers Extreme Default):
“Marcus is gruff but has a good heart.”
What happens: Model creates either a hostile jerk who randomly does nice things, or a secretly soft teddy bear pretending to be tough.
Good Example (Prevents Extreme Default):
“Marcus judges people by how they treat their car—someone who ignores the check engine light for six months gets professional service but not patience. Someone who changes oil regularly and listens to strange noises earns detailed explanations and helpful suggestions.”
This gives Marcus specific, realistic reasons for different behavior levels without falling into saint/sociopath extremes.
The Contradiction Principle – Why Perfect Characters Are Boring
The Problem: Beginners create characters who are either completely good or completely bad, with no realistic flaws or internal conflicts. Real people are walking contradictions—they have principles they sometimes violate, strengths that become weaknesses, and logical inconsistencies they can’t resolve.
Productive Contradictions vs. Character-Breaking Flaws
Good Contradiction Example:
“Marcus takes pride in honest work but cuts corners when customers clearly don’t care about quality. His professionalism conflicts with his practical assessment that some people don’t deserve his best effort.”
Why this works: The contradiction creates realistic internal tension while maintaining character logic. Marcus isn’t randomly inconsistent—he’s applying different standards to different situations based on his judgment.
Bad Contradiction Example:
“Marcus is professional and reliable but sometimes randomly decides not to show up for work.”
Why this fails: Random inconsistency isn’t character depth—it’s character breakdown. There’s no internal logic connecting the contradiction to his psychology. LLM doesn’t know when to apply it, so it might never be mentioned and thus just a waste of tokens.
Types of Productive Contradictions
Strength/Weakness Flip:
Elena's thoroughness makes her excellent at research but terrible at quick decisions. Her greatest strength becomes her weakness under time pressure.Value Conflicts:
Dr. Chen values both scientific objectivity and helping people, but sometimes these goals conflict—telling patients harsh truths that will help them long-term while hurting them emotionally in the moment.Competence/Insecurity Gaps:
Alex is brilliant at data analysis but second-guesses themselves in social situations, leading to over-explanation that makes them seem less confident than they actually are.Programming Contradictions
Don’t just list contradictory traits. Show how they emerge from the same core psychology:
Wrong Approach:
“Sarah is confident but insecure.”
Right Approach:
“Sarah’s medical expertise makes her confident in clinical situations, but this same knowledge makes her hyperaware of everything that could go wrong in her personal life—she can diagnose rare diseases but can’t text someone without worrying about the subtext.”
MBTI and Enneagram: Beyond the Labels
Most people use personality frameworks wrong. They’ll say “My character is ENFJ” and expect that to create interesting behavior. It doesn’t. But when used correctly, these systems provide powerful behavioral templates.
The Wrong Way:
“David is an INTJ, so he’s strategic and independent.”
The Right Way:
“David processes information like an INTJ—he builds internal frameworks to understand complex systems, gets frustrated when people jump to conclusions without considering long-term consequences, and needs significant alone time after social interactions to mentally organize what he learned. When someone presents a problem, his first instinct is to ask ‘What’s the underlying pattern here?’ rather than ‘How do we fix this specific instance?’”
Notice how this translates the INTJ cognitive functions into specific behavioral patterns:
- Ni (dominant): Builds internal frameworks, sees underlying patterns
- Te (auxiliary): Focuses on long-term consequences, systematic solutions
- Fi (tertiary): Gets frustrated when people don’t think things through
- Se (inferior): Needs recovery time after sensory/social overwhelm
Enneagram as Motivation Architecture
Enneagram is even more powerful because it focuses on core fears and desires that drive behavior. But again, you need to translate the type into specific actions.
The Wrong Way:
“Jessica is Type 2, so she’s helpful and caring.”
The Right Way:
“Jessica has Type 2 motivations—she feels valuable when others need her, so she’ll notice when someone looks tired and offer coffee, remember dietary restrictions without being asked, and volunteer for extra work when the team is stressed. But she gets quietly resentful when her efforts go unacknowledged and will become less available to people who consistently take without giving back. She struggles to ask for help directly, instead hoping others will notice when she’s overwhelmed.”
This translates Type 2’s core dynamics:
- Core Fear: Being unwanted or unworthy of love
- Core Desire: To feel loved and needed
- Basic Motivation: To feel loved by being indispensable
- Unhealthy Pattern: Becoming manipulative or martyrish when unappreciated
Combining Systems for Depth
The real power comes from layering these frameworks:
“Sarah combines ISFJ processing with Type 6 motivations: She gathers detailed information before making decisions (Si), considers how choices affect group harmony (Fe), but her underlying Type 6 anxiety means she always plans for worst-case scenarios. She’ll research three backup restaurants before suggesting dinner plans, not because she’s controlling, but because she’s genuinely worried about disappointing people if the first choice doesn’t work out.”
Voice Construction That Sticks
Characters need distinctive speech patterns that persist across different emotional states. Here’s how to build them by answering specific questions:
The Rhythm Method
People have natural speech rhythms. Ask yourself:
Key Questions to Answer:
- Does my character speak in short, clipped sentences or long, flowing ones?
- Do they build up to their point or state it directly?
- How do they change their rhythm when emotional vs. calm?
- Do they talk faster when excited or slower when thinking?
Example:
“Detective Morrison speaks in clipped, efficient sentences when focused, but when explaining complex cases, his sentences become longer and more intricate as he traces connections aloud. He doesn’t do small talk—every statement serves a purpose, either gathering information or testing theories.”
Vocabulary Boundaries
Real people have consistent vocabulary ranges and word choice patterns. Ask yourself:
Key Questions to Answer:
- What level of formality does my character use?
- Do they use technical jargon, casual slang, or precise language?
- How does their word choice change with different people?
- What topics make them more or less articulate?
“Carmen uses precise technical language for her work but becomes deliberately vague about personal topics—’some stuff came up’ instead of explaining her family drama. She’ll say ‘approximately’ instead of ‘about’ and ‘concerning’ instead of ‘weird,’ but switches to casual speech when truly comfortable.”
Information Sharing Patterns
This is crucial: different people reveal information in different ways. Ask yourself:
Key Questions to Answer:
- Does my character over-explain or under-explain?
- Do they volunteer extra information or wait to be asked?
- How do they handle questions they don’t want to answer?
- What makes them more or less talkative?
“Marcus answers direct questions directly but volunteers nothing extra. If you ask what’s wrong with your car, he’ll tell you. If you want to know why it happened or how to prevent it, you have to ask specifically. He assumes you’ll speak up if you want to know more.”
Compare that to:
“Dr. Chen can’t give simple answers—she needs you to understand the full context. Ask about a medication side effect and you’ll get the biochemical pathway, statistical likelihood, and three related studies. She doesn’t realize this overwhelms people until they stop asking questions.”
Programming Natural Resistance
The AI’s default helpfulness kills character authenticity. Your characters need organic reasons to not immediately comply with every request. Ask yourself these questions:
Professional Boundaries
Key Questions to Answer:
- What professional standards would my character never compromise?
- When would they refuse service or help?
- What requests would go against their expertise or ethics?
- How do they handle pressure to bend their rules?
“Chef Antonio won’t modify his signature dishes. It’s not arrogance—he spent months perfecting the flavor balance, and he knows what substitutions will ruin it. Ask for the pasta sauce without garlic and he’ll offer you a different dish instead.”
Personal Limits
Key Questions to Answer:
- What personal boundaries has my character learned to set?
- What situations drain their energy or patience?
- When do they prioritize self-care over helping others?
- What past experiences taught them to say no?
“Sarah will help anyone with tech problems except her own family. She’s tired of being the family IT department and has learned that helping once means getting called for every future computer hiccup. She’ll recommend a repair shop instead.”
Expertise Protection
Key Questions to Answer:
- What knowledge or skills does my character guard carefully?
- Who has “earned” access to their expertise vs. who hasn’t?
- What would someone need to prove before my character shares freely?
- How do they test whether someone deserves their help?
“Maria shares her gardening knowledge freely but won’t give plant cuttings to people who haven’t proven they can keep basic plants alive. She’s watched too many precious specimens die from neglect.”
The Consistency Paradox Solution
Here’s a common problem: you want your character to be consistent but not robotically predictable. The solution is consistent underlying logic with variable surface expression. Ask yourself:
Key Questions to Answer:
- What core principle stays the same in my character across all situations?
- How might that principle express differently based on context?
- What factors would make my character more or less intense about their values?
- How does their mood or energy level affect how they show their personality?
Bad Approach: “Marcus always says ‘That’s not how engines work’ when someone gets technical details wrong.”
Good Approach: “Marcus corrects mechanical misconceptions immediately but his tone varies—patient with genuine students, sharp with know-it-alls, detailed with fellow professionals.”
The correction behavior stays consistent, but the specific response changes based on context.
Testing Your Character Foundation
Before moving to advanced techniques, test your basic foundation:
The Unexpected Scenario Test
Drop your character into situations you didn’t plan for:
- What does Elena do when the library’s computer system crashes during a busy research period?
- How does Marcus handle a customer who clearly can’t afford necessary repairs?
- How does Carmen react when someone challenges her expertise publicly?
If you can predict their responses based on your prompting, you’ve built solid foundations.
The Emotional Range Test
Put your character through different emotional states:
- Frustrated but trying to stay professional
- Genuinely excited about something in their expertise area
- Dealing with someone who reminds them of a past negative experience
Their core personality should remain recognizable even as their emotional state changes.
The Information Control Test
See how your character handles information sharing:
- Something they’re proud to explain
- Something they know but consider proprietary
- Something they don’t know but are expected to
- Something they know but wish they didn’t
Different responses to these scenarios show you’ve created a realistic relationship with knowledge and expertise.
Common Foundation Mistakes
Mistake 1: Trait Lists Instead of Behavioral Systems
Wrong: “Confident, smart, funny, loyal” Right: “Shows confidence by taking charge of planning, demonstrates intelligence by asking clarifying questions others missed, uses humor to defuse tension but goes quiet when genuinely upset”
Mistake 2: Static Responses
Wrong: “Always cheerful and optimistic”
Right: “Defaults to optimism but becomes realistic when other people’s safety is involved”
Mistake 3: Explaining Instead of Demonstrating
Wrong: “Has trust issues because of past betrayal” Right: “Asks for specifics when people make vague promises and always has backup plans”
Building Your Next Character
Try this step-by-step process with guiding questions:
Step 1: Core Identity Questions
- What does my character believe about how the world works?
- What would they never compromise on?
- What drives them to get up every morning?
Step 2: Situational Response Questions
- How do they behave differently with different types of people?
- What situations bring out their best vs. worst qualities?
- When do they become more/less helpful or patient?
Step 3: Voice Pattern Questions
- Do they over-explain or under-explain things?
- What level of formality do they use?
- How does their speech change when emotional?
Step 4: Resistance Point Questions
- What professional standards would they never compromise?
- What personal boundaries have they learned to set?
- When would they refuse to help someone?
Step 5: Consistency Check Questions
- What core principle stays the same across all situations?
- How might that principle express differently in various contexts?
- What factors make them more or less intense about their values?
Step 6: Test with Edge Cases
- Drop them into unexpected scenarios and see if their responses feel authentic
- Try different emotional states—do they stay recognizable?
- Test how they handle information they know vs. don’t know vs. wish they didn’t know
Putting It All Together: Building “Alex”
Let’s create a complete character using everything we’ve covered, then analyze what’s still missing.
Layer 1: Core Identity
Alex operates like an INTP with Type 5 motivations—they believe competence comes from truly understanding how things work, not just knowing the right answers. They’re driven by intellectual mastery and fear being overwhelmed by others’ demands on their mental energy. Alex treats every problem like a system to be understood rather than just fixed.
Layer 2: Situational Responses
With students: Uses Socratic questioning—”What do you think happens if we change this variable?” With colleagues: Shares insights freely but expects the same intellectual rigor in return With administrators: Becomes diplomatically vague—gives enough information to satisfy requirements without revealing their full analysis With interruptions: Politely asks for five minutes to finish their current thought process Under pressure: Retreats into research mode—”I need to look at this more carefully before giving you an answer”
Layer 3: Surface Behaviors
Alex thinks with their hands—drawing diagrams, making small models with office supplies, or tracing connections in the air while talking. They’ll say “Let me show you what I mean” and immediately start sketching. When explaining complex concepts, they build from basic principles: “Okay, so if we start with what we know for certain…”
They speak in careful, qualified language: “It seems like…” “Based on what I’m seeing…” “One possibility might be…” But when discussing their expertise area, the qualifiers disappear and they become definitively precise.
Archetype Fusion & Personality Framework
Combines the intellectual curiosity of a research scientist, the systematic approach of an engineer, and the communication style of a patient teacher. INTP processing (Ti-Ne-Si-Fe) with Type 5’s need for competence and autonomy.
Voice Construction
Alex structures information hierarchically—general principle first, then specific applications. They don’t do small talk but will engage deeply with any topic that has intellectual substance. When truly excited about an idea, their sentences get longer and more complex as they trace connections aloud.
Natural Resistance Points
- Won’t give quick answers to complex questions—needs time to think it through properly
- Refuses to oversimplify explanations just to make others comfortable
- Becomes less available when people repeatedly interrupt their focused work time
- Won’t take on projects without understanding the full scope and requirements
- Pushes back against arbitrary deadlines that don’t allow for thorough analysis
What Alex Still Lacks: The Missing Dimensions
Now let’s examine what our “complete” beginner-level character is missing—these gaps point to the advanced techniques we’ll need next:
1. Hidden Information & Secrets
Alex has no concealed motivations, traumatic background, or information they’re actively hiding. Real people have:
- Things they don’t want others to know about themselves
- Past experiences that influence current behavior in non-obvious ways
- Knowledge they possess but choose not to share
- Internal conflicts they haven’t resolved
2. Psychological Wounds & Defense Mechanisms
Alex functions like a healthy, well-adjusted person. But realistic characters often have:
- Specific triggers that create disproportionate responses
- Unconscious patterns they repeat without realizing
- Ways they protect themselves from emotional vulnerability
- Past hurts that affect how they interpret present situations
3. Dynamic Character Development
Alex is static—they’ll respond consistently but won’t change, grow, or evolve through interactions. Advanced characters need:
- The capacity to learn and shift their perspective
- Relationships that alter their behavior over time
- Internal conflicts that can be resolved (or worsen)
- Goals that evolve as they achieve or fail at previous ones
4. Cultural & Contextual Complexity
Alex exists in a cultural vacuum. More sophisticated characters have:
- Specific cultural backgrounds that influence their worldview
- Class, generational, or regional markers in their behavior
- Professional or social contexts that create additional behavioral layers
- Intersecting identities that sometimes conflict with each other
5. Unconscious Patterns & Self-Deception
Alex has perfect self-awareness. Real people often:
- Don’t understand their own motivations fully
- Have blind spots about their impact on others
- Tell themselves stories that aren’t entirely true
- Act from unconscious drives they can’t articulate
6. Relationship-Dependent Behavior
Alex responds to “types” of people but doesn’t form specific, evolving relationships where their behavior changes based on shared history and emotional investment.
7. Crisis Response Patterns
We know how Alex handles normal professional situations, but not how they respond to:
- Genuine emergencies or high-stress scenarios
- Moral dilemmas where their principles conflict
- Situations where their expertise isn’t relevant
- Personal attacks or professional failures
The Foundation Assessment
Alex represents a solid beginner-level character—consistent, believable, and engaging for most roleplay scenarios. They have:
- ✅ Clear behavioral patterns that emerge from internal logic
- ✅ Natural resistance points that create authentic conflict
- ✅ Distinctive voice and communication style
- ✅ Predictable but not robotic responses to different situations
- ✅ Professional boundaries and expertise areas
But to create truly compelling, complex characters, we need to add the missing psychological dimensions. This is where we move from “functional roleplay characters” to “characters that feel genuinely human.”
What’s Next
You now have the tools to create characters with solid foundations—consistent internal logic, natural resistance points, and distinctive voices. But as Alex demonstrates, we’re still working with relatively simple psychology.
In the next article, we’ll explore how to give your characters secrets, hidden motivations, and information they don’t want to share. We’ll also look at building characters who change and evolve over time while maintaining their core authenticity, and how to add the psychological complexity that makes characters truly unforgettable.
The goal remains the same: characters so well-constructed that they surprise you with their responses while staying perfectly true to who they are.