Seite - 200 - in Critical Issues in Science, Technology and Society Studies - Conference Proceedings of the 17th STS Conference Graz 2018

scored on an 11-point scale, zero representing no influence and ten representing absolute act of eating as a result of the factor. Using results of the survey, we structured an agent-based model in Netlogo (Netlogo, n.d.; also see Railsback & Grimm, 2012). Agents walked along a one-mile stretch of street. (See figure 1.) One agent parameter was walking speed, which was random-normally distributed Fig. 1. Pedestrians on street: Portion of model world (eaters on a slower distribution than noneaters) based on a survey of the literature (see Bosina & Weidmann, 2017; Johansson & Kretz, 2012; Sharbafi & Seyfarth, 2017; Wagnild & Wall-Scheffler, 2013). Pedestrian density was derived by combining data from three United States cities (Chicago, 2008; Minneapolis, 2009; Tempe, 2015). Another set of agent parameters was a score on each theoretical factor. Scores on factors taken from the survey were aggregated into an indicator and each agent acquired own scores-on- factors as state variables assigned within a random-normal distribution around the indicator (parameters: mean, standard deviation). If an agent’s composite score (likelihood-to-eat score) generated by these factors exceeded a threshold, then the agent became an eater. If the threshold was not exceeded, the agent maintained a noneater status and continued on its way. In order to derive the threshold, we first conducted a one-mile visual observational field run on a business/commercial urban street and recorded proportion of eaters against noneaters. Then, taking likelihood-to-eat scores from survey data, the threshold was derived as the point which split survey respondents into the same proportion of eaters and noneaters as the field-obtained proportion. An emergent condition was built around social facilitation. Encounter with an eater and repeated encounters with eaters gradually increased the noneater’s likelihood-to-eat score. Some of these agents finally crossed the threshold after as many as nearly three-dozen encounters which were either distinct, continuous by virtue of consistent proximity over time or both distinct and continuous. Even with so many encounters, many an agent continued to remain a non-eater, as its likelihood-to-eat score never crossed the critical threshold. In the current paper, we wish to compare results, using a different method of agent parameter estimation. In the previous analysis, means of factors were used to define center of the distribution for random-normal assignment of scores-on-factors. In the current study, we are using multiple regression to estimate the center. The reasoning is encapsulated in the argument that factors tend to work in consonance with one another. A regression equation distributes factor weights based on consideration of factors involved. The reader should be aware that we are not using multiple regression here to generate a prediction model. We are using it as a rational means for articulating a more networked, wholistic expression of the phenomenon. 200