Trust-Based Human-Machine Collaboration Mechanism for Predicting Crimes

Abstract

In today’s world, with the shifting nature of artificial intelligence (AI) to explainable AI, which involves humans and machines working and complementing each other, there is a need for a mechanism to govern their collaboration. We have proposed a trust-based mechanism to manage collaboration between them. Our trust-based mechanism has the ability to quantify human trust into a mathematical model. The proposed trust-based framework will facilitate decision making when humans and machines are involved in a process. This framework will ensure that either of them never under trust or over trust each other by computing trust information based on their history. To validate our proposed framework, experiments are performed on Indianapolis Crime Data which contains actual crime information, machine predictions, and police feedback. Results have shown that how the trust of both entities can impact the decision making of the police towards machine predictions.