Irrationality and Indecision

Today's post is by Jan-Paul Sandmann (Harvard University), on his recent paper "Irrationality and Indecision" (Synthese, 2023).

Jan-Paul Sandmann

What is wrong with preferring some option a to option b, b to c, but c again to a? Why shouldn’t one cling on to such cyclical preferences? The standard response is that one ought not hold on to cyclical preferences because one could be money pumped as a result. By focusing on the binary comparisons alone, it would seem reasonable to pay some amount of money for b rather than c, a rather than b, as well as c rather than a. 

If the agent however takes these actions, she ends up with the option she started with, c, while having paid some money. And that clearly does not seem sensible: acting upon a preference cycle would not be in the agent’s interest. The money pump argument thus draws one to conclude that the agent should get rid of her cyclical preferences. 

The argument is powerful, but it also makes some constraining assumptions. Cyclical preferences lead to a self-defeating situation just in case the agent acts while holding on to these preferences. Moreover, when concluding that the self-defeating action implies that the person should give up on the preferences, we assume that only outcomes matter to the agent’s evaluation of a given choice.

In my paper, "Irrationality and Indecision", I suggest a different way of looking at an agent who holds cyclical preferences. Rather than being straightforwardly irrational, such preferences could indicate that the agent entertains conflicting views about some options. If the agent, moreover, deems these views to be important for justifying her choice, and upon further consideration fails to resolve the cycle, then her preferences ground indecision rather than irrationality.  

Yet, one may ask, are the irrationality and indecision interpretations necessarily irreconcilable? The answer I give is: “no, they need not be.” It is perfectly conceivable that the agent ends up indecisive because of underlying irrationality. And so, instead of drawing a clear conceptual wedge between irrationality and indecision, I hope to show that viewing cycles through the lens of indecision offers a valuable explanatory alternative. 

When deeming someone to be undecided because of a preference cycle some non-instrumental consideration come into play: An agent may be unwilling to choose on the basis of the preference cycle independent from the outcomes to which her choices lead. These justificatory considerations, moreover, matter from the first-personal viewpoint of the agent, rather than from that of a third-personal adviser (a perspective the money-pump argument encourages one to take on). 

Tying cycles to indecision also changes our normative evaluation of an agent. In some cases, it may be defensible to cycle among some options if doing so brings out the salient aspects of a choice to oneself. In particular, if one resolves the corresponding indecision not merely by a coin toss, but by further deliberation about the options, then preference cycles could benefit agents. Insofar as cyclical preferences provoke one to revise or reinforce one’s views, they are crucial aspects of developing standards to justify one’s choices. As such, they are an important part of becoming a reasoning agent.