Over time, the loss of human expertise through the use of AI can compromise the quality of the AI itself, and at worst, imperceptibly. This is the result of a new study by researchers from the University of Passau and Arizona State University, recently published in the prestigious journal Academy of Management Review.
Image: Google DeepMind / Pexels
Tech companies around the world are making headlines for job cuts, many citing the increased use of artificial intelligence (AI). A new study by researchers at the University of Passau and Arizona State University in the US shows that such decisions may be short-sighted.
Knowledge learned by machines can help organizations accumulate experience, such as when experienced employees leave the company. However, machine learning models are based on past training data and become outdated when the reality they represent changes. The need to update these models and the resulting “use and repair” cycle poses risks to businesses.
In their study “Fading Memory: The Role of Machine Learning in Organizational Knowledge Depreciation,” Professor Sin Gerlach of the University of Passau and Professor Don Lange of Arizona State University show how companies can get caught up in this cycle.
AI will take over. While AI systems perform tasks such as quality control in manufacturing processes, employees use their relevant expertise less often, forget about it, or leave the company altogether.
knowledge loss occurs: In this way, when tasks are handed over to AI, human expertise is lost and new employees have less expertise to gain.
AI models become obsolete: Older AI models need to be updated with new training data, model variables need to be checked for relevance, or incorrect predictions made by the model need to be evaluated by experts. Such updates require human expertise, which is increasingly scarce.
This means that AI models can become knowledge traps as they get older. “The loss of human expertise can undermine the quality of AI models over time, potentially in a creeping and unnoticed way,” warns Passau-based information systems expert Professor Sin Gerlach. “Human knowledge can be further undermined when employees uncritically internalize the predictions and decisions of outdated AI models.” In the study, the authors emphasize that long-term use of AI can only be successful if companies simultaneously maintain human expertise.
“Our findings highlight the need to maintain human expertise, otherwise the effective use of AI systems may be jeopardized over time.” – Professor Sin Gerlach, University of Passau.
Effective use of AI requires human expertise
“Our findings show that the use of AI in organizations has long-term, unintended consequences,” says Professor Gerlach. “They emphasize the need to retain human expertise; otherwise, over time, the effective use of AI systems may be jeopardized.” According to the researchers, the companies’ decisions mentioned in the introduction can be viewed differently. This is because they ignore human know-how, which will weaken the AI in the long run.
The study was recently published online in the Academy of Management Review, one of the most prestigious scientific journals in the field of management. This research is a conceptual theory study, not based on new empirical data, but combines existing findings from organizational research and computer science to form a new model. The authors derive a process theory that explains how and why a phenomenon, in this case the loss of organizational knowledge due to AI, occurs.
About the author
The lead author is Professor Jin Gerlach, Head of the Department of Data and Information Management at the University of Passau. His research focuses on technology-driven changes in business and society. Co-author Professor Donald Lang researches and teaches ethical issues in management at Arizona State University’s WP Carey School of Business.
The two authors met during Professor Lange’s research visit to the University of Passau. He spent one year as a Mercator Fellow in the Passau DFG Research Training Group 2720 “Digital Platform Ecosystems” investigating the interaction between public value and digital platforms. Mercator Fellowships are funds awarded by the German Research Foundation (DFG) to researchers who are intensively involved in projects.
This text has been machine translated from German.
[ad_2]
Source link
