"Clementine Will Remember That": Methods to Establish Design Conventions for Video Game Narrative

This paper argues that narrative game design should be treated as an empirical research area rather than a field that borrows its craft from film, television, and literature. Its central proposal is the establishment of design conventions: concrete, transferable methods that reliably produce predictable effects on audiences. The authors distinguish these from abstract design patterns, which resist measurement, and from broader design principles, such as Murray's notion of "scripting the interactor." To establish conventions, they propose a four-phase approach that moves from identifying candidate techniques in acclaimed narrative games through focus groups and review analysis, to validating them through controlled quantitative testing with established user-experience scales.

A recurring theme is methodological rigor. The authors caution that frequently used techniques may be design fads rather than effective conventions, and they stress that typical samples of 25 to 30 participants per condition are underpowered for the small-to-medium effects likely to be found. They recommend larger samples, a priori power calculations using GPower, and consistent reporting of effect sizes such as Cohen's d. Validity concerns are also addressed, including control of online study conditions, random assignment, prior gaming experience, gender balance, and the limitations of relying heavily on convenience samples of university students.

The early empirical results illustrate both the promise and the limits of this program. An online study with 60 participants found that introductory scripting which over-promises player control reduced character identification, while precise scripting increased perceived agency. A follow-up with up to one hour of play of Firewatch confirmed that agency scripting raised perceived autonomy. A physiological study of Life is Strange using skin conductance and heart rate, however, with only 17 participants, produced no significant findings, which the authors attribute largely to insufficient sample size. The authors conclude by calling for transparent reporting of study limitations, pre-registered power analyses, and further research into how conventions vary across contexts and extend beyond games.