Human Trust Factors in Image Analysis

Abstract

Advancements in robotics and machine learning technologies have increased the prevalence of human-machine interactions and collaborations in the workplace. Several studies have identified trust as a major factor in how efficiently human-machine interactions occur and in how errors are recognized and handled. Little work has been done to identify how this human-machine trust compares to human-human trust, and how an individual’s preference for human-sourced information may interfere with their human-machine relationships, and vice versa. Outside the workplace, people consume media that has become saturated by altered and out-of-context imagery. Thus, our ability to evaluate the veracity of graphical information has been compromised. Our experiment seeks to identify factors of implicit bias in how humans analyze information when it comes from a machine (algorithm), or from a human (subject-area expert). Our results highlight the need for developing a cultural computational literacy.