Coordinating Human and Machine Learning
Coordinating human and machine learning for effective organizational learning (Sturm et al. 2021)
Machine Learning Overview:
Involves deriving patterns from data using learning algorithms for real-world problem-solving.
Common types include supervised, unsupervised, and reinforcement learning.
Hybrid approaches combining human and machine learning are discussed.
Organizational Learning:
Organizational learning relies on a complex system of interacting learners.
Human learning is a social process, while machine learning is based on data processing.
Exploration and exploitation are key concepts influencing belief diversity within organizations.
Role of IT:
IT can facilitate or hinder knowledge creation, application, and dissemination.
IT is not just a tool for human learning; it can also support machine learning.
Components of Machine Learning:
Involves a narrow, real-world problem, learning algorithms (e.g., neural networks), and hyper-parameters.
Human setup is crucial; poorly done setups can introduce biased ideas into the system.
Reconfiguration is needed to align machine learning systems with current human problem understanding.
Learning Myopia and Human Preferences:
Humans naturally prefer known solutions over new, unproven ideas (learning myopia).
Machine learning acts as a remedial factor, allowing humans to learn according to their preferences.
Machine learning systems' capabilities can free up time for human activities.
Effectiveness of Organizational Learning:
Reconfiguration is essential for aligning machine learning with human understanding.
Human exploration is time-consuming, while machine learning swiftly discovers patterns.
Turbulent environments can render organizational knowledge obsolete.
Coordination of Human and Machine Learning:
Human exploration is initially superior, but machine learning agents equalize learning rates.
Machine learning agents with low initial learning capability depend on human exploration for diversity.
Machine learning agents with high initial learning capability reduce the need for human exploration without sacrificing long-term effectiveness.
Reconfiguration Intensity:
Higher reconfiguration intensity leads to higher long-term organizational knowledge.
Human learning behavior moderates the effects of reconfiguration intensity on organizational learning effectiveness.
In turbulent environments, human exploration is crucial for long-term organizational learning.
Considerations for Reconfiguration:
Reconfiguration is required to acquire high levels of organizational knowledge.
Humans should not be excluded from the loop, and domain expertise is crucial.
Moderate reconfiguration is recommended; excessive reconfiguration can be harmful.