Automatic Reverse Engineering of Human Beavior Based on Text for Knowledge Acquisition


We propose a novel approach for building cognitive architectures based on Wisdom of Crowd. As knowledge needed for an intelligent system is difficult to gather and heavily depends on a programmer's bias, we decided to automize the process of surveying reactions, decisions, opinions, etc. For demonstrating the usability of our approach we implemented Web-mining functions into our system and tested it by confronting with over 100 ethically-significant real world problems, e.g. "killing a man", "stealing money", "bribing someone", "helping people" or "saving environment". The accuracy was 86% when used emotional consequences discovery, however we show how adding social consequences discovery, natural language processing tools and bigger data sets are able to refine the results and increase correctness of moral judgment. We discuss the importance of such algorithm based on crowd behavior, which opens a path for retrieving universal transcultural rules when the system becomes multilingual and compares results from different nations.

Back to Table of Contents