Modelling Familiarity for Intelligent Personalized Social Mobilization

Zhengxiang PanAAAI 2017 · DOI

Introduction

Social mobilization—the large-scale coordination of individuals to achieve scientific, social, or political goals—is a cornerstone of modern crowdsourcing. However, current AI-driven approaches often suffer from "local trapping," where information fails to diffuse across distant locations, and an over-reliance on a few highly connected individuals. This paper addresses these gaps by integrating human factors, specifically the psychological state of participants, into the mobilization process. By focusing on the role of familiarity, the research aims to move beyond simple task allocation toward "superlinear productivity" in collaborative environments.

The Role of Familiarity

A major limitation in current algorithmic approaches is the assumption that tasks are completed by individuals in isolation, ignoring how a user’s psychological state affects their performance. This research identifies familiarity—how well a person knows a task or environment—as a critical but overlooked variable. The paper adopts a three-stage model of familiarity:

  1. Cognitive Stage: The initial acquisition of basic knowledge.
  2. Associative Stage: The transition toward process-based performance.
  3. Autonomous Stage: The ability to execute actions smoothly with minimal cognitive effort.

Preliminary results suggest that higher perceived familiarity does more than just increase speed; it enhances the perceived usefulness of the system and significantly boosts the likelihood of long-term user adoption.

Methodology: The CrowdAsm Framework

To operationalize these insights, the author introduces CrowdAsm, a novel dynamic team-formation algorithm designed for complex tasks requiring diverse skill sets. Unlike standard models that assign tasks to individuals, CrowdAsm enables the real-time formation of teams with complementary skills.

The algorithm utilizes a Lyapunov network optimization framework to balance competing constraints, including quality, time, and budget. Rigorous theoretical analysis demonstrates that CrowdAsm is asymptotically optimal, meaning it identifies the most efficient team configurations provided that participants follow the system's recommendations.

Empirical Foundation: The Pokémon GO Study

The framework is grounded in empirical data from a large-scale survey of over 300 Pokémon GO players in Singapore. Pokémon GO was selected as a proxy for social mobilization because it successfully combines gamification, time-critical incentives, social interaction, and location awareness.

The resulting dataset includes:

  • Demographics: Ethnic identity and background within a multi-racial population.
  • Psychological Profiles: Data from the Big Five personality traits (10-question test) and affective disposition (20-question test).
  • Behavioral Patterns: Time and money spent, willingness to collaborate, and prior familiarity with the franchise.

Key Insights and Results

The research moves the needle from individualistic crowdsourcing toward collaborative crowdsourcing. By leveraging human factors like emotion, trust, and familiarity, the goal is to achieve productivity that is greater than the sum of individual efforts.

Key findings include:

  • Performance Multiplication: Familiarity correlates strongly with better user performance and higher system adoption rates.
  • Theoretical Optimality: The CrowdAsm algorithm provides a mathematically sound basis for managing complex, real-time social coordination.
  • Dataset Contribution: The survey represents one of the first large-scale efforts to map how personality and background influence responses to technology-powered mobilization.

Limitations and Future Directions

The author acknowledges the risk of survival bias, as studying a successful instance of mobilization like Pokémon GO may not fully explain why other efforts fail. Furthermore, while CrowdAsm is theoretically sound, its real-world efficacy is still being tested.

Future work focuses on the development of a collaborative crowdsourcing mobile application to validate these models in real-world scenarios. This platform will integrate the computational model of familiarity to refine how teams are formed and tasks are personalized, ultimately seeking to prove the framework's ability to drive superlinear productivity in the field.

Summary updated

BibTeX
@inproceedings{pan2017modelling,
  author = {Zhengxiang Pan},
  title = {{Modelling Familiarity for Intelligent Personalized Social Mobilization}},
  booktitle = {AAAI Conference on Artificial Intelligence},
  year = {2017},
  doi = {10.1609/aaai.v31i1.10521},
}
Figure