Games people will play: a step toward development of a method to assess probability of play for instructional games

Abstract

The question of what makes games likely to be played was addressed through exploratory multivariate analyses. Undergraduate students played and evaluated ten games on a Likert questionnaire. Five objectives were set forth related to needs for design or evaluation and improvement of games which might then be used as teaching tools. The first objective was to develop an interim criterion which would reflect probability of play. Twelve items collected by interviewing game players were reduced to six in number, based upon ability to discriminate between games in this study. These six items produced one principal component which was able to separate several of the games along one dimension. The variate was highly correlated with the items "fun" and "interesting." Objective two was to evaluate items reported by Neideffer and Evans (1981) and Malone (1980) to be elements of games related to their acceptance. The 22 items were reduced to nine in number via regression while retaining most of the 46% of the criterion variance accounted for by the entire 22 item set. While response means indicated that all items were elements of games, the proportion of variance accounted for shows that the entire domain of necessary elements was not represented. The third objective was to develop a method to determine the dimensions of games which lead to a high probability of play. Principal components analysis was used to reduce the nine predictors to four variates. Multidimensional scaling produced two readily interpretable dimensions, one separating joystick games from keyboard games and a second separating games having constant motion from those which do not. These dimensions were highly intercorrelated with the principal component variates, suggesting that related items might contribute reliable variance to the principal components. The fourth objective was to determine if any subgroups of players existed in the selected sample of players for which game design modification might be required to maintain play. Q-factor analysis, inter-correlation of participants with an average pseudo-participant, and plots of MDS weights revealed no subgroups in the present sample. The possibility of finding subgroups with other games and items or in other samples was not ruled out. The final objective was to develop a regression equation to predict probability of play as reflected by the interim criterion. Regression weights for the PCA variates were determined which accounted for 40% of the criterion variance. Pseudo-replications of the regression analysis revealed the weights to be unstable beyond the first two principal components.