It’s been a really long time since I last wrote one of these posts. Honestly, I barely did any further research on my topic during the break, and I’m not going to pretend otherwise. But the time off wasn’t entirely wasted. In the context of Design & Research 2, I had to do a more structured reflection on the work I’ve already done – and that turned out to be surprisingly useful. Not because it revealed anything radically new, but because it gave me some distance. And distance, it turns out, is exactly what you need to notice the things that have been quietly bothering you for a while.
So before I move on to my plan for learning a new skill this semester, I want to take one more look at where my research actually stands – what I built, what I think works, and what I’d do differently if I were starting today.
What I’ve Built So Far
For anyone who hasn’t read the previous posts: I’ve been working on a systematic framework for analyzing fictional antagonists and antiheroes. The goal is to create a method that’s fast enough to scale across large numbers of characters and rigorous enough to produce results you can actually compare statistically.
The framework has four layers:
Observable Traits – the character as an artefact. Drawing from the ACIS framework by Linke and Prommer, this layer looks at what you can directly see and hear on screen: screentime, spoken dialogue, visual identity, movement, and physical presence. No outside context, no assumptions – only what the viewer actually experiences.
Personality Profile – the character as a represented being. This is the Big Five layer. I developed a custom 20-item coding sheet (4 items per trait, rated on a 1–7 scale) that lets me score any character on Extraversion, Agreeableness, Conscientiousness, Neuroticism, and Openness – purely from on-screen behavior. To make it trustworthy, I also built in a coder agreement check: two people code the same character independently, and we measure how closely their scores correlate. Anything above 0.70 is considered acceptable.
Symbolism & Motivation – what the character represents beyond their actions. A looser, more qualitative layer that captures the central symbol, lead motif, and thematic meaning the character carries. Harder to quantify, but important for explaining why a character resonates, not just how they behave.
Creation & Cultural Context – when and by who the character was made. Demographics of creators and directors, year of release, cultural background. This layer contextualizes everything else and helps explain effects like nostalgia or the influence of the specific era a character was born in.
The theoretical backbone behind all of this is Jens Eder’s Character Clock – though I consciously dropped the Symptom dimension to keep the analysis scalable. Covering all four of Eder’s dimensions in depth per character would take forever. I’d rather analyze 100 characters at a useful level of depth than analyze 5 at an impractical one.
What I’d Actually Change
Here’s the thing that’s been nagging at me since the reflection exercise: the scope is too narrow.
My entire research focus up to this point has been on villains – or more specifically, sympathetic villains. And when I actually look at what my framework does, I realize that framing never made much sense as a hard boundary. The coding sheet doesn’t know whether it’s rating a villain or an antihero. The ACIS-based observable traits layer doesn’t care. The Big Five certainly doesn’t. These tools work on any fictional character.
Restricting the framework to villains feels kind of arbitrary at this point. More than that: it actually limits the analytical value of the whole thing. One of the main goals was always to find out what makes a sympathetic villain different from a clear-cut hero or a straightforward one-dimensional antagonist. But if I only ever analyze the villains, I have nothing to compare them against. That’s a pretty obvious problem.
What I think makes more sense going forward is to broaden the scope to fictional antagonists and morally complex characters in general and then use the framework to actually compare them across the full spectrum. Heroes, antiheroes, villain protagonists, secondary antagonists, the whole range. The patterns only become interesting when you can see the contrast.
There’s also something I want to tighten up on the qualitative side. The Symbolism & Motivation layer is the weakest link in terms of consistency. Right now it’s mostly free-form notation, which makes it almost impossible to compare across characters in any structured way. I haven’t figured out the right solution yet – whether it should be a fixed set of categories, a tagging system, something else entirely – but it’s the one part of the framework that still feels unfinished.
And finally: the coder agreement is still untested in practice. I have the method, I know the threshold, but I haven’t actually run the test with a second coder on a real sample. That’s the most important thing to validate before I start building any kind of database seriously.
So that’s where things stand. A solid structure that needs a wider lens, a qualitative layer that needs more rigidity, and one empirical test that still needs to happen.
Next up: what I’m planning to learn this semester, why I chose it, and – more interestingly – what it actually has to do with all of this.
See ya!
