Collective incentives reduce over-exploitation of social information in unconstrained human groups

Dominik Deffner, David Mezey, Benjamin Kahl, Alexander Schakowski, Paweł Romańczuk, Charley M. Wu, Ralf H. J. M. Kurvers

TL;DR

Researchers used immersive 3D virtual reality to show that while "scrounging" (following others to find rewards) benefits individuals, it often cripples group-level performance. By shifting from individual to group-based incentives, the team found that humans naturally adopt more selective social learning strategies, effectively buffering the collective against the "tragedy of the commons" in information sharing.

What Problem Does This Paper Address?

In any social group, individuals face a fundamental trade-off: do you explore the environment yourself (personal information) or do you follow the success of others (social information)? This is known as the producer-scrounger dilemma.

While social learning is a powerful shortcut to success, relying on it too heavily creates a "tragedy of the commons." If everyone scrounges, no one is exploring, and the group fails to discover new resources. Most previous research on this has used static, simplified simulations or highly constrained laboratory tasks. This paper tackles the challenge of understanding these dynamics in unconstrained, naturalistic human groups—where movement, visual perspective, and spatial constraints all influence decision-making in real-time.

Methods

The authors developed a naturalistic foraging task in a 3D immersive environment (built in Unity). Groups of four participants searched for "coins" using virtual metal detectors.

The study’s technical centerpiece is the Social Hidden Markov Decision (SHMD) model. Because researchers cannot "see" a participant's intent, they used this unsupervised model to infer latent behavioral states from high-resolution movement data (25Hz). The model categorizes behavior into:

  1. Individual Exploration: Searching for patches independently.
  2. Social Relocation: Actively moving toward a successful peer.
  3. Exploitation: Extracting rewards from a discovered patch.

By analyzing turning angles, changes in distance, and relative bearings, the SHMD model identifies when a participant has "switched" their strategy based on what they can see in their limited field of view.

Datasets

The researchers generated a novel high-resolution dataset consisting of:

  • 160 participants (40 groups of 4) performing the collective task.
  • 40 solitary foragers acting as a control group to establish a baseline for individual performance.
  • High-resolution time-series data: X-Y coordinates, orientation vectors, velocity, and "visual social information" (exactly what each player could see of their teammates at any given millisecond).

The data and the Unity-based experimental framework are available on GitHub and Zenodo.

Key Insights and Ideas

  • The Incentive Shift: When paid based on individual success, participants over-exploited social information in "concentrated" environments (few rich patches). When switched to group-level incentives, participants didn't stop using social info; they became more selective. They only followed others when the distance was short or the "scrounger density" at a patch was low.
  • Individual Gain vs. Collective Loss: In concentrated environments, high scrounging rates predicted higher individual coin counts but lower group totals. This confirms that what is "rational" for a person can be "maladaptive" for the collective.
  • Spatio-temporal Trade-offs: Using time-lagged Gaussian process regressions, the authors found that being close to others is a double-edged sword. In the short term (<15 seconds), proximity helps the group exploit a found patch. In the long term, however, staying close reduces the group's "search area," leading to fewer new discoveries.
  • Calibration Over Time: Group-incentivized participants "tuned" their strategies over the course of the experiment, becoming more sensitive to the costs of crowding, whereas individually-incentivized players remained indiscriminately social.

Key Results

  • Performance Gap: Individually-incentivized groups performed significantly worse in concentrated environments compared to distributed ones (a deficit of -8.4 coins). Group-incentivized teams showed no such performance drop.
  • Selectivity: Group incentives promoted higher selectivity. For example, the probability of switching to "Social Relocation" decreased as the number of players already at a patch increased—but only for those in the group-incentive condition.
  • Discovery Rates: Solitary foragers actually outperformed groups in terms of total coins collected, highlighting the inherent "interference" costs of social foraging in competitive environments.

Limitations and Future Work

The authors acknowledge several areas for expansion:

  • Group Size: The study was limited to groups of four. Collective dynamics often change non-linearly as group size increases.
  • Zero-Sum Dynamics: The current task featured "exploitative" competition (once a coin is taken, it's gone). Future work could explore "non-zero-sum" scenarios where social information use might be more universally beneficial.
  • Fixed Environments: The patches were static in quality. Real-world environments often involve mobile resources or patches that deplete and regenerate at different rates, which would add another layer of complexity to the producer-scrounger trade-off.

Summary updated

Abstract

Collective dynamics emerge from countless individual decisions. Yet, we poorly understand the cognitive processes governing dynamically-interacting individuals in human collectives under realistic conditions. We present a naturalistic immersive-reality experiment where groups of participants searched for rewards in different environments, studying how individuals weigh personal and social information and how this shapes individual and collective outcomes. Capturing high-resolution visual-spatial data, behavioral analyses revealed individual-level gains -but group-level losses- of high social information use and spatial proximity in environments with concentrated (vs. distributed) resources. Incentivizing participants at the group (vs. individual) level facilitated adaptation to concentrated environments, buffering excessive scrounging. To infer discrete choices from unconstrained interactions and uncover the underlying decision mechanisms, we developed an unsupervised Social Hidden Markov Decision model. Computational results showed that participants were more sensitive to social information in concentrated environments frequently switching to a `social relocation' state where they approach successful group members. Group-level incentives reduced participants' overall responsiveness to social information and promoted higher selectivity over time. Finally, mapping group-level spatio-temporal dynamics through time-lagged regressions revealed a collective exploration-exploitation trade-off across different timescales. Our study unravels the processes linking individual-level strategies to emerging collective dynamics, and provides new tools to investigate decision-making in freely-interacting collectives.
BibTeX
@inproceedings{deffner2024collective,
  author = {Dominik Deffner and David Mezey and Benjamin Kahl and Alexander Schakowski and Paweł Romańczuk and Charley M. Wu and Ralf H. J. M. Kurvers},
  title = {{Collective incentives reduce over-exploitation of social information in unconstrained human groups}},
  booktitle = {Nature Communications},
  year = {2024},
  doi = {10.1038/s41467-024-47010-3},
}

Loading related content...

Figure