From Santa Fe Institute Events Wiki

Last Tree Project


This is a group project of the SFI, lead by Juan Camilo Cardenas, Charles Efferson, and John H. Miller. Please do not cite or use any of the material contained herein, without the express written permission of the authors. Thanks.

Next Steps

  • JCC will look at his current data and identify (near) extinction events, and see if there are any insights to be had for this project.

The project site where the data will come is at: There are at least two sources of such data from human behavior: -Field experiments with villagers in colombia and thailand using the "forest game" (see a first papr reporting data from these games at: )

-Lab experiments using the "pacman" game of forest extraction with students forraging over a space of trees with regrowth

  • All three will continue to refine and expand this wiki with the near term goals of
    • Having a good sense of the basic problems and likely classes of explanation
    • Creating a set of simple experiments to explore the above
    • Figuring out a simple pilot experiment
    • Funding (we will need some $ for running the experiments---probably not that much money in the end)
      • Home institutions?
      • SFI International Program?


Throughout history we see humans driving various systems to extinction (trees on Easter Island [Diamond, Collapse], essentially American Bison, etc., etc.). Can a set of clever experiments illuminate the causes of such behavior?

Suggested Hypotheses for Why Things Get Driven to Extinction

Don't anticipate it (Diamond, ch 14)

Don't perceive it once it occurs (Diamond, Ch 14)

Fail to solve it once it is perceived (Diamond, ch 14)

Social dilemmas

Group think failure


  1. Overconfidence on the system's capacity to recover.
  2. A lack of economic (value) importance of the resource in period t or any t+n.


The general experimental framework will be the harvesting of a renewable resource in a very simple framework. For example, there are X trees in the world, at each round you can harvest some of the trees for experimental earnings, and at the end of each round we add some trees as a function of how many trees remain. We should probably have a fixed probability of the experiment ending at the conclusion of each round, so that we can impute a discount rate. The experiment ends when the system goes extinct.

The main observables will be total harvest and time to extinction.

Experimental Conditions

  1. Individual versus group decision making
    What happens when one individual is in charge of all of the decisions versus a group of individuals? Note that by "group" we could mean a group of individuals a'la the usual public goods experiments or a "governmental" group where the individuals have to agree with one another prior to making a decision. It would be interesting if we found conditions under which an individual drives the system to extinction, as that would imply that social dilemmas are not necessary for such events
  2. Underlying growth processes
    How does the structure of the underlying growth process impact extinction? Are there some forms that are much harder for decision makers? The experimental literature has largely ignored dynamical decision-making in a CPR setting. Nonetheless, harvesting is an intrinsically dynamical process. Treating them as such in experiments would be one important way into the problem. There is some literature, though. Here's a paper:
    1. Types of growth processes. We need to figure out a few, simple, prototypical growth processes from ecology. At the moment we have three characteristic processes discussed here:
    2. Are these the full set of prototypical growth processes? Answer: Probably not the full set, but they represent three very different and widely discussed postulates.
    3. What if we put lags in the growth process? Answer: Discrete-time models/experiments involve an implicit lag. Or is something else meant here?
  3. Abstract versus real objects
    Are people better at dealing with these process when they are abstract (e.g. tokens) or real (e.g. trees or some other resource with which they have some actual experience)?
  4. Change in scale
    Can we induce a sense of "false abundance" by simply changing the scale of the problem (let every tree become 10 shrubs), and does this have any implication for behavior?
  5. Information and uncertainty impact this system?
    1. Does noise in the growth process make this easier or harder? (Similar to lags)
    2. Does noise or uncertainty in the actions of others in a group make extinction more likely?
    3. Can we induce "moral wiggle room" by, for example, having some probability that at any time, harvesting trees will help the ecosystem and increase growth, or that the growth is not tied to the existing number of trees? (This latter option is also known as the "bad and inconclusive science" approach a'la Bush---there is a chance that we have no impact on global warming...) The wiggle room would be more interesting if they could, for free, know what kind of world they are operating in. Presumably, under such a condition, you would want to know as it would help you maximize payoffs, but if you wanted the wiggle room, you might not want it revealed and just act as if it was the more favorable world.
  6. Discount rates
    Presumably, under high discount rates, extinction is the preferred option. Does the discount rate matter to the participants? Can we make them stay in the game too long?