The article examines the Hot Stove Effect, a psychological bias affecting learning from experiences. It highlights how negative past outcomes skew an individual's ability to accurately assess potential opportunities. Specifically, the effect describes a tendency to steer clear of alternatives deemed less favorable, leading to persistent underestimation of their value. The authors extend the theory beyond typical avoidance behavior, establishing that even moderate sampling reduces exposure enough to maintain underestimation errors over time. Additionally, they present findings pertaining to Bayesian learners, who similarly tend to underappreciate expected values of alternatives due to a negativity bias in their learning process.
The 'hot stove effect' points to a bias in learning algorithms where negative past experiences lead to a tendency to sample less from less favorable alternatives, allowing errors of underestimation to persist.
When alternatives yield negative estimates, learners have a reduced inclination to sample from these options, perpetuating a cycle of misjudgment and hindering the correction of over- and underestimation biases.
Collection
[
|
...
]