Abstract
This research addresses the pervasive issue of non-compliance with Information Security (IS) policies in organizations and explores innovative solutions using Artificial Intelligence (AI) and Machine Learning (ML) algorithms. The study focuses on developing an AI-driven artefact through crowdsourcing to autonomously detect and modify human behaviour, particularly in response to security breaches and policy violations. The research employs the Design Science Research (DSR) approach, integrating crowdsourcing, design thinking, and the scientific method in developing the AI algorithm.
The study's contribution lies in providing practical solutions for IS practitioners, leveraging AI to proactively monitor and modify human behaviour. The proposed AI artefact aims to streamline early detection and modification of unsafe behaviour, enhancing compliance with IS policies. The research methodology involves crowdsourcing ML algorithms, applying the Theory of Planned Behaviour (TPB) as a theoretical framework, and employing DSR principles.
Results demonstrate the development and evaluation of Algorithm 2; a binary classification model designed to predict and modify poor user behaviour related to information security policies. The model underwent iterative improvements based on participant feedback on GitHub, showcasing effectiveness through a loss value of 0.1376. Evaluation metrics, including accuracy, precision, recall, AUC-ROC, and MSE, provide insights into the model's classification accuracy and error measurement.
The study's practical importance lies in the integration of AI and crowdsourcing for behaviour monitoring and modification, addressing challenges in user attitudes, subjective norms, and perceived behavioural control. It contributes to the field by exploring the synergy of AI and crowdsourcing, automating behaviour modification, and creating an AI-driven tool for behaviour monitoring. The implications for stakeholders range from researchers finding new directions to technology companies capitalizing on behaviour-
vi
tracking solutions. The study concludes with an acknowledgement of limitations, proposing future research areas, and summarizing its valuable contribution to understanding and addressing poor user behaviour through AI-driven solutions.
Keywords: Artificial Intelligence, User Behaviour, Theory of Planned Behaviour, Crowdsourcing, Algorithm, Artefact.