Exploring behavior patterns in human and machine interactions

A framework based on response-dependence theory is proposed to analyze human and machine interaction patterns. The interaction is an information processing and exchanging system with agents responding to signals from the decision perspective on multiple channels. Agents' pattern differences may emerge via a multiple-input multiple-output model on the sequential machine data. The connectivity with different decision-making factors embodies the interactive narrative features. The magnitude of the frequency response is compared in a parallel analysis. The interaction relationship will emerge in frequency response comparison. In the empirical study, we compare AlphaGo and human professional players' behaviors in the game of Go. We find a similar connectivity structure in vertical analysis and coincidences with the interaction relationship in the game. However, the response magnitude interval differs between AlphaGo and human professional players in several stimulus-response pairs. Integrated with static comparison, we find AlphaGo is more sensitive to long-term payoff changes than human professional players. The framework and empirical studies indicate that theory and techniques from cross-disciplinary can provide a perceptive and objective explanation of behavior patterns in human and machine interaction. The proposed framework might benefit scientists in research on AI ethics and machine behavior interpretation with big data techniques.