Social Behavior Bias and Knowledge Management Optimization

Yaniv Altshuler, Alex Pentland, Goren GordonSBP 2015 · DOI

TL;DR

This paper proposes that humans, unlike rodents, are suboptimal at managing information flow, leading to inefficient decision-making in social networks. By analyzing trading data, they demonstrate an inverted-U relationship between information sources and financial gains, suggesting that optimizing this "information temperature" could improve network performance.

What Problem Does This Paper Address?

In today's information-saturated world, individuals and systems are constantly bombarded with data. While having information is crucial for effective decision-making, there's a critical point where too much information becomes detrimental, hindering performance due to processing constraints. This phenomenon, often described as "information overload," is particularly relevant in social networks where information flows between individuals, influencing collective outcomes. The paper addresses the gap in understanding how humans, specifically in economic contexts, manage this information flow compared to other biological agents and how this impacts their decision-making efficiency and the overall performance of social systems.

Methods

The authors adopt a "neuro-social-physics" integrated approach. They leverage findings from neuroscience on how biological agents process information, particularly the concept of "novelty management" and the trade-off between too little and too much information, which results in an inverted-U efficiency curve. This principle is then applied to a large-scale social trading platform. They analyze the statistical properties of information flow within this network and model the dynamics of social network information flow based on individual novelty management principles. The core idea is to treat individual decision-makers as "atoms" in a system, where their interactions create network-level dynamics akin to statistical physics ensembles.

Datasets

The primary dataset used is from "eToro," an online social financial trading platform. This platform provides financial transaction data for foreign exchanges, equity indices, and commodities. It includes data from approximately 3 million registered accounts and over 40 million trades conducted between 2011 and 2014. This dataset is described as a "highly detailed large financial trading database."

Key Insights and Ideas

  • Human Suboptimality in Information Management: Unlike rodents, which appear to have evolved to maintain an optimal signal-to-noise ratio (SNR) for information processing, humans tend to be suboptimal. This means humans don't naturally operate at the peak of the inverted-U curve for information efficiency.
  • The "Wrong Thing" Maximization: The paper hypothesizes that humans, in their pursuit of maximizing monetary gains, often disregard the principles of information management. They may not internalize that excessive information can be detrimental, leading to confusion and suboptimal decisions.
  • Social Network Dynamics and Information Flow: The "spread" of individuals along the inverted-U information management axis within a social network can lead to unstable and inefficient macro-scale dynamics. This suggests that the collective performance of a network is highly sensitive to how its members manage information.
  • "Information Temperature" as a Control Parameter: The authors draw an analogy to physical systems, suggesting that the "temperature" of the network—representing the diversity of information spread—can be manipulated. Raising the temperature could create a more diverse spread, while lowering it could lead to a more homogenous network, potentially tuning the system towards optimal information management.

Key Results

The study demonstrates an observable inverted-U relationship between the average financial gain of investors and the number of information sources they utilize. This finding validates the hypothesis that there is an optimal number of information sources for maximizing returns, and that both too little and too much information are detrimental. The paper also shows that the distribution of people along this inverted-U curve leads to unstable network performance, which is further validated by a global inverted U-shape observed in macro-scale network performance.

Limitations and Future Work

The authors acknowledge that their work is a conjecture and a first step in a larger research program. They suggest that changing the distribution of people's positions along the information management axis can significantly impact network performance. Future work could explore specific manipulations, such as adjusting the "temperature" of the system (i.e., the diversity of information spread) to achieve a more optimal information management regime. The paper focuses on financial decision-making, and its findings might need to be generalized to other domains of social decision-making.

Summary updated

BibTeX
@inproceedings{altshuler2015social,
  author = {Yaniv Altshuler and Alex Pentland and Goren Gordon},
  title = {{Social Behavior Bias and Knowledge Management Optimization}},
  booktitle = {SBP},
  year = {2015},
  doi = {10.1007/978-3-319-16268-3\_27},
}
Figure