Apr 17: Fairness and Uniform Risk Assessment
Category: Social Sciences
Posted by: Aaron
There is a common inclination in cases of deep uncertainty to apply probabilities uniformly across all possible future states, despite this inclination being unsupported by any theoretical principle. The notion may be derived from (i.e. be applied analogously from) the behavior rules guiding distributive justice wherein it is accepted that, if differentiating reasons are not to be found, then dividing a resource (or requirement) equally among possible recipients is prescribed. The rule internalized for the historically common cases of distributive justice in past human society may be providing the uniform probability distribution intuition in the decision calculation procedure. This case might therefore be an instance of the evolutionary history of humans generating a reliable bias in modern human decision problems and thus merits consideration for institution design.
Regardless of the details, it is a reasonable bit of speculative evolutionary history to think that in proto human societies there were many occasions where distributive justice played a key role in assigning rewards and penalties to the members of social groups. When it comes to distributing a resource (or a liability) there are many reasonable options for assigning each relevant person their share. These include distribution according to contribution, merit, need, status, efficiency, value, and others. These different distribution schemes require a fair amount of information about each individual's contribution, merit, etc. The method that gets used may depend on culture, existing institutions, environmental conditions, and the resource to be distributed, but the point I'm making here is that sometimes these decisions had to be made without the relevant information. In cases where the most relevant kind of information is unavailable but another sort is, a group may simply substitute a scheme despite some degree of injustice that results. However, in cases where a) insufficient information exists for any scheme, b) the distribution institution is inflexible regarding the scheme used, and/or c) significant disagreement exists regarding the appropriate scheme, there must be available an alternative "fallback" scheme.
Throughout human evolution I suspect that different groups have experimented with a great variety of fallback schemes...i.e. schemes that require only a bare minimum of information to implement. But one scheme that seems to have been frequently and independently hit upon is uniform distribution. To give each member of a group an equal share of a resource one only needs to know the total resource value and the number of group members. Who counts as a group member (e.g. individual, family, household, etc.) may involve some dispute due to its affect on the actual distribution, but there would have been cases where group membership was unambiguous (e.g. hunting parties, festival dancers, unwed males, etc.). In order to minimize disputes over resource distribution in such cases an equal share scheme seems a plausible mechanism. Over the long period of evolution such mechanisms seem to have been adopted as fairness norms to preserve social justice. These are thick terms, but exactly these sorts of reactions have been consistently uncovered in many social primate studies (and additionally citing their link to our evolutionary past).
Whether or not "equal shares" is a valid fairness norm for cases of resource distribution with incomplete information is not at issue here. The point is that humans and other primates do, in fact have these intuitions and reactions in the relevant cases. The evolutionary explanation is only helpful, no matter how plausible, if it can be tested and used to improve decision making in the future. The problem is not that there is something better to do in cases of deep uncertainty than uniform distribution, it's just that there isn't any reason to do any particular thing. For distribution it's not the case that the equality scheme assumes each member is equal, only that there is no sufficient consideration that distinguishes them. But it can be harmful to group decision processes if people insist that equal distribution is the correct decision making procedure and fight the inclusion of other considerations.
As a solution, my thinking is that if we can manipulate a decision environment so that the decision making under uncertainty looks minimally similar to a case of distributive justice then the intuition of equal distribution as best practice would be proportionally weakened. The cases are different, but since there really just isn't anything "correct" to do for cases of uncertainty – and that is hard for humans to cope with – what we end up with is reliable bias that we can adapt our institutions to account for and avoid. How precisely to do that will need to be determined application by application, but realizing that there is something to be done here is the first step. The descriptive import is to correctly explain the behaviors observed in terms of adaptive processes and to predict how people will react in the relevant contexts. The primary normative import is to defuse the false rationalization of uniform distribution of probability under uncertainty. But this idea, if defensible, may also contribute to refining our concept of fairness, to focusing our criterion for uncertainty, and to seeing how evolutionary explanations can provide practical guidance for modern institutional design.
Regardless of the details, it is a reasonable bit of speculative evolutionary history to think that in proto human societies there were many occasions where distributive justice played a key role in assigning rewards and penalties to the members of social groups. When it comes to distributing a resource (or a liability) there are many reasonable options for assigning each relevant person their share. These include distribution according to contribution, merit, need, status, efficiency, value, and others. These different distribution schemes require a fair amount of information about each individual's contribution, merit, etc. The method that gets used may depend on culture, existing institutions, environmental conditions, and the resource to be distributed, but the point I'm making here is that sometimes these decisions had to be made without the relevant information. In cases where the most relevant kind of information is unavailable but another sort is, a group may simply substitute a scheme despite some degree of injustice that results. However, in cases where a) insufficient information exists for any scheme, b) the distribution institution is inflexible regarding the scheme used, and/or c) significant disagreement exists regarding the appropriate scheme, there must be available an alternative "fallback" scheme.
Throughout human evolution I suspect that different groups have experimented with a great variety of fallback schemes...i.e. schemes that require only a bare minimum of information to implement. But one scheme that seems to have been frequently and independently hit upon is uniform distribution. To give each member of a group an equal share of a resource one only needs to know the total resource value and the number of group members. Who counts as a group member (e.g. individual, family, household, etc.) may involve some dispute due to its affect on the actual distribution, but there would have been cases where group membership was unambiguous (e.g. hunting parties, festival dancers, unwed males, etc.). In order to minimize disputes over resource distribution in such cases an equal share scheme seems a plausible mechanism. Over the long period of evolution such mechanisms seem to have been adopted as fairness norms to preserve social justice. These are thick terms, but exactly these sorts of reactions have been consistently uncovered in many social primate studies (and additionally citing their link to our evolutionary past).
Whether or not "equal shares" is a valid fairness norm for cases of resource distribution with incomplete information is not at issue here. The point is that humans and other primates do, in fact have these intuitions and reactions in the relevant cases. The evolutionary explanation is only helpful, no matter how plausible, if it can be tested and used to improve decision making in the future. The problem is not that there is something better to do in cases of deep uncertainty than uniform distribution, it's just that there isn't any reason to do any particular thing. For distribution it's not the case that the equality scheme assumes each member is equal, only that there is no sufficient consideration that distinguishes them. But it can be harmful to group decision processes if people insist that equal distribution is the correct decision making procedure and fight the inclusion of other considerations.
As a solution, my thinking is that if we can manipulate a decision environment so that the decision making under uncertainty looks minimally similar to a case of distributive justice then the intuition of equal distribution as best practice would be proportionally weakened. The cases are different, but since there really just isn't anything "correct" to do for cases of uncertainty – and that is hard for humans to cope with – what we end up with is reliable bias that we can adapt our institutions to account for and avoid. How precisely to do that will need to be determined application by application, but realizing that there is something to be done here is the first step. The descriptive import is to correctly explain the behaviors observed in terms of adaptive processes and to predict how people will react in the relevant contexts. The primary normative import is to defuse the false rationalization of uniform distribution of probability under uncertainty. But this idea, if defensible, may also contribute to refining our concept of fairness, to focusing our criterion for uncertainty, and to seeing how evolutionary explanations can provide practical guidance for modern institutional design.
