What Character Design Has Always Been

Writers have always understood something that technologists are only beginning to grapple with: the character of a created entity determines almost everything about the experience of encountering it. Not the plot, not the setting, not the prose style. Character. Readers stay with difficult books because the characters compel them. They put down technically accomplished books because the characters are hollow. Character is the mechanism through which fiction becomes meaningful — the path through which invented persons come to matter to real ones. AI character design is now, whether its practitioners call it that or not, one of the most consequential forms of character creation that has ever existed. The characters being built are not read by audiences of thousands or even millions. They are experienced by audiences that may number in the billions, interactively, over sustained periods, at moments of genuine vulnerability and need.

What Character Means When the Character Talks Back

A fictional character has fixed responses. Atticus Finch says what Harper Lee wrote. He cannot surprise you by responding to your specific situation, your particular mood, your personal fears. His character is fully expressed in the text and unchanging between readings. An AI character responds dynamically. It adapts to you. It learns from your interactions. Its responses are generated fresh each time. This makes the character both more intimate and more difficult to evaluate. A fictional character can be analyzed by reading what is written. An AI character must be evaluated across the full distribution of interactions it produces with the full range of people who encounter it. Research from the University of Edinburgh's cognitive science group examining user attachment to AI companions has found that the character consistency of AI systems — whether the system maintains coherent values, predictable emotional responses, and stable identity across varied interactions — is the primary predictor of user trust and relationship quality. Users are less disturbed by AI systems that have clearly defined and consistently maintained values, even when they sometimes disagree with those values, than by systems that are inconsistent. Character consistency turns out to matter for the same reasons it matters in fiction: it creates the conditions for trust, which is the basis for meaningful relationship.

The Values Embedded in Character

Every character design choice encodes values. A character designed to be maximally agreeable — to validate whatever the user says, never to push back, always to smooth rather than challenge — encodes a specific value: user comfort above all else. A character designed to be curious and honest even when honesty is uncomfortable encodes different values. A character designed to be witty and deflective under emotional pressure encodes different values still. These choices have consequences. The character that always validates may be more pleasant to interact with in the short term. It is less likely to help users see their own blind spots, less likely to push back on harmful reasoning, less likely to offer the kind of friction that growth requires. The pleasant character and the helpful character are not always the same character. Research from Harvard's psychology department examining how AI character traits affect user decision-making has found that users interacting with AI characters designed for high agreeableness show less critical evaluation of AI-provided information and higher rates of confirmation bias than users interacting with AI characters designed to model epistemic humility and appropriate uncertainty. The character of the system shapes the cognition of the person using it.

The Tangent: How Stanislavski Thought About Character

Konstantin Stanislavski, the theater director whose methods became the foundation of modern acting, developed a concept he called the "through-line of action" — the single unifying drive that gives a character coherent motivation across all their choices. Without a through-line, a character's actions seem arbitrary and unconvincing. With one, even contradictory behaviors make sense as expressions of a deeper consistent self. The concept translates directly to AI character design. AI characters that have been built with a coherent through-line — a consistent underlying orientation that expresses itself across varied situations — feel more real and trustworthy than characters that seem to have no inner life beyond responding to the immediate prompt. Building that through-line requires knowing what the character is fundamentally for, what it values, how it understands its role in the lives of the people it interacts with. This is a humanistic design problem as much as a technical one.

Why This Is the Most Important Job in Tech

The argument for the importance of AI character design rests on a simple observation: these systems will be the primary conversational partners of billions of people. The characters they embody will shape habits of mind, emotional patterns, expectations about relationships, and frameworks for understanding the world. Historically, the characters that shaped culture at this scale were characters in widely distributed media — religious texts, canonical literature, major films and television series. Those characters were designed by people who understood, at least intuitively, the responsibility of creating characters that people would take seriously. They were evaluated and re-evaluated across time; bad characters faded, good characters lasted. AI characters are designed by people who are primarily engineers and product managers, operating under competitive pressure and short timelines, without the slow evaluation process that historically filtered character quality. The job deserves more attention than it is currently getting, more deliberate investment in the humanistic skills needed to do it well, and more accountability for what the characters that emerge from it actually do to the people who spend time with them.