Predicting personality associations evoked by multicolored appearance of virtual agents: an exploratory study

Designing agents' color appearance that can evoke desired color-personality associations (CPAs) is still a challenge. To respond to the challenge, this study proposed a multicolor personality index (MCPI) model to predict the CPAs evoked by agents using the color-personality indices for single colors together with the histogram of agents' image. Specifically, the color personality index for each bin of an agent's histogram was first calculated based on the color-personality indices for single colors; then, the predicted values of CPAs were obtained by calculating the weighted average of color-personality indices of all bins in the histogram. To improve the MCPI model, a simple color-attention index model was proposed to evaluate the effects of color on attention and the effects were introduced into the MCPI model, that is, the multi-color personality index based on color attention (MCPI-CA) model. To validate the two models, an experiment was conducted to collect the CPAs of five traits (i.e., extraverted-introverted, moody-unemotional, agreeable-disagreeable, organized-disorganized, and wide interests-narrow interests) evoked by the multi-colored appearance of agents. The results showed that (a) observers in general perceived the CPAs evoked by color appearance in similar way; (b) for most agents, the CPAs evoked by multi-colored appearance could be predicted well by the MCPI model; and (c) the MCPI-CA model usually had better performance than the MCPI model due to the inclusion of the effects of color attributes on attention. In addition, the limitations of the proposed models and their implications for designers were also discussed.