Artificial Intelligence Bias Minimization Via Random Sampling Technique of Adversary Data
Muath A. Obaidat, Nila Singh, George Vergara · 2021
Artificial Intelligence is a growing field in technology that mimics the human neural network in order to deduct patterns based on specific datasets. Unlike conventional methods of programming where the code is told explicit rules, AI uses data to predict processes. However, due to AI's prediction of future behavior, it is highly susceptible to data tampering from adversaries who may flood the program with false information. Previous solutions have utilized random sampling, active learning, blockchain and human interaction in order to solve AI bias. In this paper we propose a scheme to address the AI bias by using a method of random sampling in order to mitigate the destruction done to hacked systems while maintaining prediction reliability.