Neural Networks and Human-Machine Interactions
Photo by Josh Riemer on Unsplash
The use of human and machine thinking in a dynamic system can potentially increase the capabilities of the system. With the use of machine learning and neural networks from the computer, the human can utilize large amounts of data, based on the output from the machine. The output from the machine, however, can be affected by the input of the data and the algorithms that are run by the machine system. According to Géron (2019), a toddler can be shown an apple and told that the object is an apple, and the toddler will understand what an apple is from there. A machine cannot have one instance to learn what an object is. A machine learning system can require large portions of data to simply understand what an object is (Géron, 2019).
A potential human factors issue that can arise from the use of neural networks and human-machine interactions is the potential for a machine to output inappropriate data. Inappropriate data can be created from multiple concepts like an insufficient quantity of training data, unreasonable effectiveness of data, nonrepresentative training data, and overfitting the training data (Géron, 2019). These concepts show that although a machine may have large quantities of data to learn from, the data itself may cause the machine to misunderstand the intentions behind the project. This can cause a negative output of information to the human that is using this data.
By combining machines and humans, there must be trust in the machine to provide relevant and timely data for the human to use. In the event that a human is using the machine's data to supplement or enhance their own performance or decision making, a human factors problem exists that may be bad output machine data and bad input human understanding of the machine’s data. The neural network in the human brain does not require the large amount of data that the machine requires to understand what an object is or how to complete a task.
According to Goldstein (2018), the human brain uses experience to build neurons in the brain to create memories and an understanding of what is happening to recall later. This recall is what allows humans to use a previous experience or encounter to understand what an object is or how it works, similar to the apple and the toddler. The ability of the human brain to quickly understand an experience and use it later in life is the difference between human learning and machine learning.
Studying Human Factors
A study must be conducted to understand how this can affect the dynamic use of humans and machines for a project and verify the validity of their use in a dynamic system. My research question in this study is, "Does human-machine interaction degrade human capabilities in a combined system." The research study would determine the validity of the output of a machine to the human operator to ensure the human was not incapable of performing their job in the event that the machine provides poor or bad quality information to the human operator.
To study the effects of a combined human-machine system, the researchers would utilize three groups: one group of human-machine operators, one group of only human operators, and one group of multiple human operators. All of these groups would perform the same tasks and would be provided the same relevant information to perform the tasks. The study would use the groups to determine if the human-machine group was able to outperform the other two groups in the study. If the human-machine group does not outperform the other group, the data will allow researchers to improve the human-machine interactions to ensure they were appropriately using the data and the machine in the process to ensure an adequate and intuitive system has been created.
According to Dolgov & Hottman (2011), research in the use of autonomous systems with human pilots showed a distrust for the autonomous systems when they performed incorrectly. This degraded the human pilot's ability to perform their job due to the continuous monitoring of the autonomous system, which doesn't exist in a system where the human is the continuous pilot. The difference is the human pilot knew what action they were taking, and the autonomous aircraft operator was relying on the machine to perform as it was tasked. If the three groups in the study follow a similar path in the research by Dolgov & Hottman (2011), the human factors issue that may arise is poor performance or bad output from the machine to the human operator.
References
Dolgov, I., & Hottman, S. B. (2011). 11 Human Factors in Unmanned Aircraft Systems. UNMANNED AIRCRAFT SYSTEMS, 165.
Géron, A. (2019). Hands-On Machine Learning with Scikit-Learn, Keras, and TensorFlow: Concepts, Tools, and Techniques to Build Intelligent Systems. O'Reilly Media.
Goldstein, E. B. (2018). Cognitive psychology: Connecting mind, research, and everyday experience. Nelson Education.