Character Likability Explained
A primer on how to think about measuring character likability in a way that can yield audience and story insights.
My guiding question is how agents and studio teams can select, create, and market better series and movies with story metrics by offering guidance on research questions they are already asking. The things I am saying in this substack are aimed at fixing mistakes I’ve heard such practitioners make when asking questions to analysts who use story metrics. Much of the time, the conceptual mistakes being made by practitioners are only obvious when researchers start trying to measure things with numbers. People get away with using words loosely until numbers get involved. This is especially true when it comes to character likability in stories.
Entertainment insights vendors claim — with some validity — they can measure character likability to answer questions such as whether to greenlight a script, franchise a character, market a series with a character, or pitch a character to high-profile acting talent.1 There are numerous well-known books that talk about writing, acting, and filming characters, and such books have provided a loose vocabulary for storytellers to use when discussing characters and their traits. This same vocabulary gets misapplied in industry analytics and AI story metrics.
Studio People Talking About Characters
Studio people don’t have the time to think through granular questions about characters in ways that could lead to better insights from story metrics. And there's currently no specialized department to handle such research, even where data-science insights are commonplace (e.g., Apple Studios, Amazon Studios do not have a social-science branch that informs the data teams). Different projects have different objectives, whether it's winning awards, attracting talent, achieving a high return on investment, maintaining TV ratings, or other performance indicators. There is more than just likeability to consider for a character. This complexity calls for a nuanced understanding of audience judgments of characters as a prerequisite to insights from character metrics.
Contrary to this need for nuanced understandings, practitioners sometimes lump all of their character problems under the umbrella term of character likeability — expecting this umbrella term to encompass all aspects of high-quality character scripting that could achieve their business goals. This lumping of things into “likeability” has anecdotally seemed apparent to me not only with studio professionals but also with students in film and new screenwriters. To say that you want a likeable protagonist in your story is conventional wisdom. Screenwriters have confusions on this topic (e.g., see this Reddit thread). Writer Scott Burns (of Bourne Ultimatum, The Informant) touched on this ambiguity in his talk at UT - Austin 10 years ago, where he reported that producers outright told him they didn’t know if his characters were likable enough to greenlight the script (revealing the simplistic, over-generalized view of likeability in major studios is at the highest levels of decision-making).
Screenwriting Literature on Characters
The screenwriting literature — taken as a whole — only helps to a certain extent. The literature prescribes a hodge-podge of different tactics but doesn’t explain how those tactics relate to specific success outcomes or the psychology of why they work. Just to check, I re-opened Save the Cat, Story, and Story Maps again for references to character traits, audience judgments, and likeability. Snyder (in Save the Cat) implies having the character do a morally admirable act near the very beginning of the story to boost likeability. He goes further by suggesting Jungian-inspired archetypes (“young man on the rise,” “good girl tempted” ), as well as to use characters that elicit audience identification, and judgments of proactivity. Sometimes he seemingly appears to be talking about audience enjoyment, but other times he is talking about a qualitatively different response like appreciation. It’s hard to tell which tactics he prescribes will work for a specific situation with a specific goal in mind. Mckee in Story similarly discusses archetypes, emotional connection, proactivity, hero’s journey, having a long emotional growth journey. Calvisi in Story Maps focuses on character motivation and action and how it yields a narrative arc with key character points, but does not go into specific measurable aspects of characters and how to get insights about characters from data. None of the aforementioned texts define what their goal is in terms of measurable success outcomes (and so they have been allowed to use language loosely). Although they acknowledge different story types with different needs, they do not connect story or character features to specific business criteria. While these three books alone have made an enormously positive contribution to storytelling, they were not intended as data-science guides for marketers or as erudite volumes of scientific theories about entertainment.
What frameworks can we use to measure things meaningfully and get such insights, though, for practical use cases? When trying to get insights out of story metrics about characters, it’s useful to have some type of systematic understanding of the functional components that underlie all of the metrics. That is, it’s nice to have a framework where we can slot all of our metrics. I’ll mention a couple of frameworks below, and what character metrics have been drawn from them.
Understanding the Connections Between Likeability and Morality for More Insightful Measurements
Quantitative entertainment theories unlock the utility of story metrics by explaining the links between all of the story metrics. Below I will mention a couple of such theories, link them together, and call out how the character metrics generated by those theories relate to marketing and development needs.
I discussed disposition theory in the last two posts and I will bring it up again here to set the stage for thinking about insights from measuring character likability as I promised in the title. When applied to TV and movies, Zillmann's disposition theory (short for disposition theory of delight and repugnance) suggests viewers become emotionally invested based on their moral judgments of characters. This morality-driven emotional attachment influences how tense or excited they feel about what happens to those characters and ultimately shapes whether they find the story satisfying, keeping them engaged throughout the narrative. Metrics like suspense, morality, enjoyment, and likability are key to disposition theory.
Audiences respond with hedonic pleasure when seeing an evil (unlikeable) character punished or a virtuous (likable) character rewarded. And they respond with repugnance to bad guys being rewarded or good guys being punished. Thinking in this basic way, we can build on this to have a framework for insight generation. For example, we can start to distinguish likeability from other factors with this framework in a way that matches the desired audience response. That is, we can label likeability as the moral appraisal of the audience. Likeability is that which makes the audience want to see a character rewarded or punished, and with what intensity. This is not a convenience label or a reductionism. This definition is consistent with loads of experiments and surveys showing how moral appraisals affect audience ratings of how much they like a character.
Hagrid is a highly likeable character because he has moral intentions.
Research pans out this idea that likeability is based on moral judgments. As a single (but notable) example among many, researchers at the Grizzard Entertainment Lab at The Ohio State University recently put all of disposition theory’s tenets into a big experiment where they tested a large cause-and-effect model. It backs up basically all of DT’s claims. In a striking confluence, perceived character morality and likeability are so tightly correlated that they collapse into a single psychological construct. Researchers sometimes call this construct character “dispositions,” but in Hollywood parlance, this is likeability. The questionnaire literally asked how much the audience “liked” the character. Moreover, I’ve looked at dozens of surveys and experiments I conducted myself, and likewise found moral factors to be the biggest predictors of likeability. Good guys are likable, bad guys are less likeable. Those are the necessary and sufficient conditions.
But aren’t some villains likable?
People in the industry often bring up the idea that villains can be likable. I’ve even heard a data vendor once mention a study where marketing researchers at Northwestern showed evidence for villains’ high likeability. (This was a common case of misreading the academics’ work.) The academic researchers were able to measure personality traits in audience members’ selves as well as personality traits in villains in an online platform for character-based content discovery called Charactour (a wonderful data source for audience judgments of characters). Upon analysis, they observed that personality and attitudinal similarity between an audience member and a villain made the audience member more likely to be a fan of that villain. Certainly, this greater fandom must be some form of appeal. The data vendor was correct to think that people like to watch stories with compelling villains. That is, she was correct that villains can make a story more appealing or be appealing themselves. But nowhere in the study do the authors claim this appeal (i.e., number of fans on the online platform) is a measure of how much the audience “likes” a character. One can see how easy it is for practitioners to use the word likeability, though, as a catch-all term for all character-quality aspects. Despite this tendency among practitioner experts, data shows audiences en masse define likeability in terms of morality.
Dear studio and screenwriting “experts” — please stop trying to tell me audiences like this guy.
There are outright challenges to this likeability and morality claim (and disposition-theory as a whole) from academics as well. Researchers from the Recreational Fear Lab mistakenly say studies that use “disposition theory assume that immoral characters could not inspire positive engagement by virtue of their very immorality.” This is a misreading of disposition theory on their part. It’s not “positive engagement” with the character DT discusses in that way. It’s likeability. The authors of that study then go on to measure participants’ moral values and correlate those moral values with the liking of villains who also have those values — in effect supporting my claim here (i.e., disposition theory) while claiming to do the opposite. Although disposition theory has major limitations (e.g., it can’t fully explain anti-hero appeal, liking and disliking might not be true opposites, most stories don’t contain pure evil or pure virtue characters, and more) the framework does give a straightforward explanation for likeability that empirically holds up. And researchers have used fancy best practices to make solid metrics for it that are useful for model training and insight automation.
Finally, when it comes to the allure of villains, things can become confusing due to levels of measurement—i.e., the intricate gradations of likeability. Far from being a binary, yes-no attribute, likeability unfurls along a complex spectrum. It's not merely a matter of black and white, but a sliding scale that stretches from the overtly dislikable to the unequivocally likable, with a middle-ground of neutrality or ambivalence. Even characters steeped in villainy can flash sporadic glimmers of likability—and conversely, heroes can exhibit flaws. These moments make us stop and think, inviting a question: are these characters in fact “likeable”? The answer hovers in the murky middle of our likeability scale, challenging us to reckon with ambiguity in a realm entertainment practitioners and data scientists often mistakenly see in binary.2
What about measuring morally ambiguous characters?
So this naturally brings us to morally ambiguous characters, such as anti-heroes. This is an important question for story developers as it relates to overall types of success outcomes the title might move the needle on (e.g., moral ambiguity usually correlates with ratings and awards whereas moral clarity with popularity and box office). I am using the term “morally ambiguous” to refer to any character who does both morally good and bad things.3 The term goes well beyond overt examples of antiheroes (e.g., Travis Bickle) to include the more mundane (e.g., Jim Halpert). When a character does both good and bad, this can provoke deliberative thought in the mind of the audience. Another form of appeal beyond likeability is how thought-provoking or interesting a character is. There are evolutionary reasons why audiences are attracted to such content. When you have a morally complex character, it becomes more important for story developers and marketers to get a nuanced understanding of exactly why that character is morally complex. For development and marketing, this helps with automated character comping and demographic targeting. Such methods can be used to help find spinoff opportunities or determine what format the character would work best in (e.g., series, limited series, movie).
Since the character is mixed, their likeability is somewhere in the middle. However, their appeal could still be high. That is, they could be compelling — worth watching — without being likable and without being deserving of reward in the mind of the audience.
The conflict between moral intuitions is what causes these slower, more complicated character judgments. This points toward our desire to measure exactly what moral values are in conflict. By having a systematic way to understand what moral values characters exhibit, we can unlock even more nuanced ways to measure them. So far we have linked morality to likeability, and likeability to overall story enjoyment. Next, we’ll break apart morality into constituent measurable components.
Entertainment researchers have primarily used two key theories to analyze morality: moral foundations theory (MFT) and, more recently, morality as cooperation (MAC). Both theories deconstruct morality into its basic elements, allowing for a detailed examination of the moral facets of characters. Each of these moral elements, which are outlined in Exhibit 1 below, can be reliably measured in scripts, despite the subjective nature of morality. By linking these elements to the concept of character likeability and disposition theory, we gain valuable insights into both storytelling and audience preferences.
Studies have demonstrated how these moral elements align with the traits of different audience segments. For example, politically conservative viewers tend to value authority, loyalty, and sanctity, while liberal viewers prioritize care and fairness. Similar trends are observed when comparing rural to urban audiences or collectivistic to individualistic communities. Variations also exist based on gender and age.
In practical terms, these insights can be applied to metrics that marketers already have available to them. We can identify the moral values most important to different audience groups. To do this, we can either use artificial intelligence for large-scale analysis or rely on consumer surveys for specific cases. This allows us to match characters with audience values, leading to more effective storytelling and targeted marketing.
Exhibit 1. A list of different measurable aspects of character morality in stories. Moral foundations theory is based on Darwinian evolution. Morality as cooperation is rooted in both evolution and game theory.
Moral Foundations Theory (MFT) - Jonathan Haidt
Care/Harm: This domain captures our concern for the well-being of others, often manifesting in compassion and nurturing.
Fairness/Cheating: This is about treating others in proportion to their actions, often linked to justice, rights, and autonomy.
Loyalty/Betrayal: Here the focus is on allegiance to one's group or community, sometimes described as in-group loyalty.
Authority/Subversion: This domain is concerned with respect for tradition and legitimate authority.
Sanctity/Degradation: This focuses on purity and sanctity, often tied to religious or spiritual beliefs.
Liberty/Oppression: This domain encompasses our concern for individual freedom and opposition to tyranny and oppression.
Morality as Cooperation (MAC) - Oliver Scott Curry
Family: This domain is focused on kin altruism, or the principle that "blood is thicker than water."
Group Loyalty: This involves mutualism, working together within a group for mutual benefit.
Reciprocity: This domain is about keeping promises, summarized as "you scratch my back, I'll scratch yours." Its focus is trust.
Heroism: This captures the essence of heroic altruism, or acts that greatly benefit others at personal risk.
Deference: This is about respect for social hierarchies and those in positions of authority.
Social Equality: This domain concerns the fair distribution of resources or benefits among members of a group.
Property Rights: This involves respect for ownership and boundaries, both material and spatial.
In-Group Love: This is about coalition building and showing preferential treatment to in-group members.
Out-Group Hostility: This domain concerns promoting one's own group at the expense of others, often tied to intergroup competition and rivalry.
As you can see, there is some overlap between MFT and MAC. But the idea here is that audiences judge characters along these criteria intuitively and often unconsciously — and measurement techniques have been developed specifically for TV and movie characters using these frameworks.
Additional Character Concepts
We’re just scratching the surface of how to think about characters, and what’s possible to scalably measure with AI or in consumer panels. I‘ve left out deliberately a thorough discussion of relatability, competence, flaws, authenticity, and several other buzzwords studios and screenwriters use. I’ll save more character stuff for future posts. But by learning about disposition theory and the morality theories above, you should be able to think about metrics tapping likeability, morality, and audiences together in a unified whole. This can clarify the types of questions you ask analysts and give you easy-to-think-through reasons why you’re investigating the question. That’s the magic of having frameworks.
It’s extremely rare for software developers to understand measurement best practices, so binary tags are the norm on platforms that can collect such data at scale. Most metrics available in entertainment lack this basic level of measurement consideration (i.e., their concepts are not binary), which removes a lot of predictive power needed for insight generation.
Oliver, M. B., Bilandzic, H., Cohen, J., Ferchaud, A., Shade, D. D., Bailey, E. J., & Yang, C. (2019). A penchant for the immoral: Implications of parasocial interaction, perceived complicity, and identification on liking of anti-heroes. Human Communication Research, 45(2), 169-201.
Eden, A., Daalmans, S., & Johnson, B. K. (2017). Morality predicts enjoyment but not appreciation of morally ambiguous characters. Media Psychology, 20(3), 349-373.