Crystal Crop Fever

How do you find the highest peak in a mountainous landscape, like the Alps? The answer seems easy: you go to the base of the highest mountain, in this case, Mont Blanc, and start climbing. But what if you did not know which was the highest mountain, and Google was not an option? Again, the answer seems trivial: you look around and find the mountain which looks the highest. You need to correct for perspective, but if you get that right, soon you are well on your way. But what if the landscape is covered in thick fog? How does one find the highest peak in a landscape, when one does not know where to look?

It turns out that a lot of the most challenging problems can be described as a landscape, where each point in the landscape represents a solution and the height of the landscape is given by how good the solution is. This is good because, in principle, if you know how to search, then you can solve ANY problem: you could use the same method you used to find the peak in the Alps to determine which solutions to test.

We use Crystal Crop Fever to study how teams search for the highest peak in problem landscapes, how we influence each other, and how we divide the tasks between us. With this research, we aim to understand the conditions in order to facilitate efficient teamwork in a general setting. 

This game helps our research in:

  1. Human problem-solving behavior in an optimization task.
  2. The conditions that affect the exploration-exploitation tradeoff.

The Science Behind

Most of the games on ScienceAtHome are a search in a complex landscape, where we look for the best solution to a given problem. This search can be conceptualized as walking through a high-dimensional complex landscape with some rationale behind what the next step should be, a process called optimization. The most efficient optimization method is very dependent on the problem type and it is not always straightforward to identify the proper method.

E.g. in quantum physics it had been shown that all peaks in the problem landscape had equal height given there were no restrictions on the properties of the solution [1]. Thus, a simple local optimization, which finds the top of the first available peak, would do just fine. However, as you introduce limitations on the solution such as how slow it is allowed to be, this assumption breaks down, and other methods are required.

There is a long tradition of social science looking into human problem solving as an optimization problem, i.e. people trying to improve their performance by paying attention to what they've done in the past, much like some computer algorithms. Yet, humans seem so much better at it (Carruthers and Stege 2013). Research of this kind has so far been purely conceptual, using optimization search as a model of how humans solve problems. Human problem solving is such a tremendously complex cognitive endeavor that even though there are strong traditions across multiple fields (psychology, cognitive science, anthropology), this research has brought little insight into how humans solve problems.

However, by creating an environment where people solve highly complex optimization problems (wrapped inside a game) we can more thoroughly understand human problem-solving behavior, by i.e. benchmarking it to computer algorithms.
Answering this question might help us to design better algorithms to solve very complicated problems like how to design a quantum computer.

In Crystal Crop Fever the landscapes are well understood and designed to test how people work together when they are confronted with an unfamiliar and complex problem. Thus, we will vary on some of the underlying conditions [2][3].