Impact of network structure on collective learning: An experimental study in a data science competition

Area of Science:

• Organizational Design
• Collective Learning
• Computational Social Science

Background:

• The prevailing theory suggests efficient communication networks enhance collective learning and innovation.
• A competing theory posits that less efficient networks may foster novel solution discovery.
• Understanding network efficiency's impact on problem-solving is crucial for optimizing group performance.

Purpose of the Study:

• To experimentally test whether efficient or inefficient communication networks lead to superior solutions in complex problem-solving.
• To compare the effectiveness of different network structures in facilitating collective learning and innovation.

Main Methods:

• Conducted 16 experimental Data Science Competitions with distributed groups of data scientists.
• Manipulated the efficiency of communication networks within these groups.
• Employed a greedy search strategy where individuals only adopted better solutions.

Main Results:

• Groups with inefficient communication networks consistently discovered better solutions than those with efficient networks.
• Inefficient networks led to higher average solution quality and superior best solutions found by group members.
• This outcome was observed across all experimental trials.

Conclusions:

• Contrary to prominent theories, inefficient communication networks appear more conducive to discovering novel and better solutions for complex problems.
• Network structure significantly impacts collective learning and the ability to find innovative solutions.
• Findings suggest a re-evaluation of optimal network designs for collaborative problem-solving in certain contexts.