-- WORK IN PROGRESS --
Here's our project leader, Charley Wu, giving a talk on our project as part of the Cognition, Collectives, and Human Culture workshop for CogSci 2020:
Producers and Scroungers
Collective search is an important and widespread phenomenon across biological systems. From bacteria to fish, birds, and mammals, individuals join forces to more effectively search for resources or safety. An important question in any social learning environment is knowing when to imitate others and when to innovate as an individual.
Producer-scrounger models divide these foragers into two distinct categories - or roles: Producers, which search for new food sources and scroungers, which join the food discoveries made by others.
Here's a classic example of this model in action made by Keynan, Ridley & Lotem (2014):
Its quite clear that this model doesn't just manifest itself among animals in the search for food, but also within many aspects of modern society, examples of which can be seen time and time again. For instance, in the world of video games, developer corporations will quickly pounce on suddenly surging trends started by someone else. The genres auf battle royale, autochess or MOBA all had their roots in minor projects made by small-time developers or hobbyists. Yet, once successful, a manifold of large-scale corporations quickly jumped on these bandwagons by releasing their own versions. (PUBG, Fortnite, TFT, Underlords, LoL, Dota2 and many others)
Individual foragers can, however assume both roles of the model at once. For instance, when releasing Heroes of the Storm, Blizzard quite clearly sought to capitalize on the recent success of the MOBA genre, yet when they released Hearthstone, they singlehandedly brought the online digital collectible card game genre (which had been a mostly niche genre until then) into the mainstream.
And indeed, a healthy balance of producers and scroungers appears to be the ideal makeup of a functional and efficient society, particularly in a corporate setting. No produces: no innovation, no scroungers: lesser quality products.
Attentional trade-offs
The trade-off between individual exploration and imitation of others is encapsulated in the producer-scrounger dilemma (Barnard & Sibly, 1981; Kurvers et al., 2009), where social imitation (scrounging) comes at the cost of reduced (attentional) resources for individual search (i.e., producing). Attempted solutions to this dilemma generally have frequency-dependent payoffs, meaning one's performance depends on the strategies used by other individuals in the population. While social learning is cheap and effective among mostly individual learners, it fails amidst an abundance of imitators, causing a collapse in both individual and group performance.
This phenomenon is known as Rogers’ (1988) paradox. However, more recent work has shown that Rogers’ para-dox can disappear in a variety of more realistic setting. For instance, when agents can flexibly switch between individual and social learning (Boyd & Richerson, 1995; Enquist & Ghirlanda, 2007; Kameda & Nakanishi, 2002), when rewards and agents are embedded in a spatial structure (Beauchamp,2008; Rendell et al., 2010; Kobayashi & Ohtsuki, 2014) or when social information is not copied verbatim, but adapted in the process of transmission (Ehn & Laland, 2012; Boyd & Richerson, 1995).
Yet the majority of these results are derived from simulations, with little experimental data about how people flexibly arbitrate between social and individual learning in realistic environments. This motivates our use of an immersive virtual environment(Fig. 1a), where a limited field of view constrains how participants allocate visual resources. In this spatially-explicit environment, visibility is naturally defined in terms of distance and orientation between participants (Fig. 1b). Thus, the environment is only partially observable at any point in time, with opportunity costs when learning either socially or individually. Visual attention allocated towards individual search detracts from social learning, and vice versa.
The Experiment
Using Minecraft, which by itself fits neatly into the producer-scrounger model, we developed an experiment for studying collective search in human groups.
Four players are placed on a field of melons (or pumpkins). When harvested, these fruits can either yield a reward or nothing at all. For melons, the rewards tend to be clustered together while with pumpkins they are scattered at random. Every time a player finds a reward a blue splash signals this to the other players.
As in natural environments, the task imposes a limited field of view, which creates a resource allocation problem: visual attention can either be allocated to focus on individual exploration or to look towards peers for social imitation. At the heart of this task is a social coordination problem, where too many imitators can cause a collapse in both individual and group fitness (Rogers, 1988).
Proceedings
Framework
(TODO: Forge mod structure and features) The vast majority of Minecraft modding is done Java and, as such, we opted for the "Forge" mod loader to do most of the heavy lifting for us.
(TODO: Visual of fireworks video and stuff)
(TODO: WebTool for starting and running etc)
(TODO: Python video thing)
Visual Field Simualation
(TODO: timemachine)