``Rather Solve the Problem from Scratch'': Gamesploring Human-Machine Collaboration for Optimizing the Debris Collection Problem

Title of the Paper

“Rather Solve the Problem from Scratch”: Gamesploring Human-Machine Collaboration for Optimizing the Debris Collection Problem

Paper Information

• Subject Area: Human-Machine Collaboration, Optimization Problems, Gamified Experiments
• Keywords: Human-Machine Collaboration, Human-in-the-Loop Optimization, Gamified Experiments, Serious Games, Post-Disaster Debris Collection Problem, Visual Optimization, Operations Research

Research Background and Problem

• Identified Challenges: Optimizing operations in critical infrastructure networks is key to mitigating the impact of disruptive events. However, traditional computational algorithms (black-box solvers) alone may fall short in addressing many complex problems. Therefore, there is a need to combine human visual capabilities with the computational efficiency of algorithms to explore the potential of human-machine collaboration.
• Significance:
  • Post-disaster debris collection is a complex societal challenge, involving a multi-objective optimization problem of assigning multiple contractors to various road segments for debris removal.
  • Designing effective optimization processes for practical applications is crucial for reducing disaster recovery costs and improving efficiency.
• Research Motivation and Related Work:
  • In traditional optimization methods, humans are often excluded, but the concept of "human-in-the-loop optimization" has gained traction in recent years.
  • Researching the role of humans in optimization problems through gamified experiments is still a nascent field. This paper focuses on studying human-machine collaboration through serious games to inspire new approaches to designing decision-support tools.

Solution

• Proposed Solution:
  • Developed a digital game called “Debris,” allowing players to solve a simulated debris collection problem in a disaster-stricken Manhattan.
  • Players can choose to construct solutions from scratch or optimize solutions based on initial algorithmic suggestions.
  • Introduced a novel interactive optimization approach that combines human visual capabilities with algorithmic collaboration.
• Innovations:
  • The gamified experiment is the first to explore the advantages and disadvantages of solving optimization problems from scratch versus starting from algorithmic initial solutions.
  • Coined the term “Gamesploring” to describe the exploration of human-machine collaboration through gaming.
  • Investigated the visual advantages of human-machine collaboration: humans can intuitively understand maps and allocate contractors effectively.
• Implementation Steps and Techniques:
  • Game Design: Developed the game interface and optimization algorithm module using Unity game engine and MATLAB.
  • The layout assigns road segments to contractors, enabling real-time algorithmic optimization and human interaction.
  • Players can use color blocks to "paint" contractor assignments or accept suggestions from the Deborah algorithm.
  • Visual tools such as heatmaps and interactive features help players better understand the problem and diagnose optimization nodes.

Research Outcomes

• Specific Findings:
  • Player-generated solutions significantly outperformed algorithmic solutions in visual optimization objectives and were comparable or superior in time and profit objectives.
  • Data revealed that humans have a significant advantage in visual tasks and can independently excel in visual objective optimization.
  • Players completed the problem in an average of 10-18 minutes, compared to the algorithm's 50 minutes of iterative computation.
• Advantages Comparison:
  • Human-machine collaboration produced more balanced solutions, demonstrating superior performance in solving multi-objective optimization problems.
  • The gamified design enhanced user engagement, transforming complex problems into a visually intuitive and highly interactive experience.
• Experimental Results and Limitations:
  • Players showed a preference for solving problems from scratch, with lower acceptance of algorithmic initial solutions (especially visually disorganized ones).
  • Data indicated individual differences related to debris distribution and player backgrounds.
  • Further research is needed to communicate the role of algorithms more effectively to enhance collaboration.
• Future Directions:
  • Explore how to design collaborative algorithms that better leverage human performance, focusing on task division and human visual preferences.
  • Expand game research to investigate human-machine collaboration across different map structures and task distributions.
  • Further explore the potential applications of gamification in real-world decision-support systems.