An Assessment of Personality in a Collective Learning System with the Five-Factor Model of Personality

摘要

This research investigates the relationship between personality traits, learning, behavior, and intelligence in a collective learning automation (CLA) embedded in a collective learning system (CLS). The FIVE-FACTOR model (FFM) of personality, a model of the major dimensions of personality, is used to assess the personality of the CLA, named Nemo, as it plays a game called Expedition. Nemo's compensation policy determines how it responds to its supervisor and how it reinforces and punishes its own behavior. Nemo's response selection policy determines how it selects its moves in the game based on its tie and reject thresholds. Nemo's intelligence is varied with its forward context: it can search the game tree of possible future responses and their expected payoffs to optimize its current response. The primary goal of this research is to measure the performance of Nemo playing Expedition. The factors are the upper and lower bounds of Nemo's compensation policy, Nemo's tie and reject thresholds, and the number of plys Nemo can search forward in the game tree. One set of performance metrics is used to measure Nemo's learning and behavior. Another set is used with observational methods to measure Nemo's FFM FACTORS and facets to assess its personality. Informal experiments are conducted, and the results show that Nemo is capable of learning to play Expedition with no forward context, and that Nemo is capable of exhibiting a range of personality with no forward context.