PhD Thesis Proposal – Li Yonggang
Title: Effect of Artificial Intelligence on Organizational Learning and Individual Knowledge Work Performance
Abstract:
This thesis examines how artificial intelligence (AI) reshapes organizational learning and knowledge work.
The first essay analyzes how organizational learning dynamics shift when human agents interact with AI systems. Drawing on and extending the exploration-exploitation framework of March (1991), I formalize an agent-based model of AI-assisted search. In this model, organizations vary in their trust in AI, the AI’s capability (in terms of generality and veracity), and the rate at which humans learn from AI outputs. The analysis reveals that AI can substitute for human exploration under appropriately calibrated levels of organizational trust, thereby altering the balance between exploration and exploitation. However, the benefits of AI are contingent on its capability; while improvements in AI veracity consistently enhance organizational learning, expanding AI influence into domains where it performs poorly can reduce long-run performance. These findings contribute to a reimagining of organizational learning in the age of AI, highlighting the consequences of integrating external algorithmic knowledge into organizational processes.
The second essay shifts the focus from organizational dynamics to individual-level cognition in knowledge-intensive tasks. In this context, AI assistance may produce paradoxical effects: enabling the rapid discovery of high-quality solutions for single tasks while simultaneously diminishing the opportunity for individuals to accumulate the cognitive knowledge necessary for sustained performance across tasks. Through controlled simulations on an extended NK fitness landscape model, I compare three search strategies: pure experiential search (local adaptation) as a baseline, cognition-based search, and AI-assisted search. The results reveal a potential tension between short-term efficiency and long-term capability development. Although AI assistance improves immediate performance, heavy reliance on it reduces the rate of experiential learning and the formation of structured mental representations of the problem landscape. This trade-off is especially pronounced in complex environments where deep cognitive representations are critical. The analysis thus identifies the conditions under which AI acts as a performance accelerator versus a substitute for cognitive development.
The two essays develop replicable computational models that illuminate how AI reshapes the mechanisms through which knowledge is generated, aggregated, and internalized. By clarifying the interplay between trust, capability, and cognitive development, this thesis advances a dynamic theory of AI-enabled organizing and offers design insights for balancing human and machine intelligence in complex problem-solving environments.