Daniel Oppenheimer 2008-09-04
From OBTnotes
- from Princeton Psych
- Speaker's site
Abstract
This paper examines the extent to which decision anomalies can be explained as an emergent property of aggregation of preferences. We assume multiple neurons or neural systems (agents) within an individual, each with different goals and preferences. We show that several puzzling decision anomalies (attraction effects, similarity effects, and compromise effects) arise as a natural consequence of aggregating these preferences. As a result "irrational" behavior at the level of the individual can possibly be explained by the summation of "rational" preferences at the neural level.
Notes
attraction, similarity, compromise
- no continuous utility function can account for all three
previous explanations
- adaptive models - different decision spaces yield different strategies
- decision field theory - sequential sampling and lateral inhibition
- leaky accumulator model - sequential sampling and assumed loss aversion
new explanation: voting architecture
- different neurons and neural systems preferentially respond to different attributes
- decision anomalies result from the aggregation across these neural influences
- different voting aggregation methods can create different outcomes given a single static set of preference profiles (e.g. plurality vs antiplurality).
- methods other than plurality that assign some weight to preferences below the favorite will result in the attraction, and compromise effects
- anything between a borda count (1,0.5,0) and antiplurality (1,1,0) result in similarity effects
simulation model:
- cobb-douglass functions in 2d attribute space to describe agent prefences
- aggregation function of form (1, 1-delta, 0) where delta is the utility distance between the first and second choice
future directions
- the model is fine with more options in the decision space
- more attributes but that will require something more complicated than cobb douglass
- different distributions of agents (using uniform now)