טכניון מכון טכנולוגי לישראל
הטכניון מכון טכנולוגי לישראל - בית הספר ללימודי מוסמכים  
Ph.D Thesis
Ph.D StudentPlonsky Ori
SubjectChoice Behavior in Unstable States
DepartmentDepartment of Industrial Engineering and Management
Supervisors Professor Ido Erev
Dr. Kinneret Tedorescu
Full Thesis textFull thesis text - English Version


Abstract

Many behavioral phenomena can be the product of a tendency to rely on small samples of past experiences. Why would small samples be used, and which experiences are likely to be included in these samples? Previous studies suggest that a cognitively efficient reliance on the most recent experiences can be very effective. Indeed, the assumption that recent outcomes have a large effect on subsequent choice underlies the most popular models of learning. For instance, most models assume that a recent favorable outcome increases the choice rate of an alternative, and the effect monotonically diminishes in time (i.e. assume a positive recency effect). In contrast, we explore a very different and more cognitively demanding process explaining the tendency to rely on small samples: exploitation of environmental regularities. We start with a computational investigation of wide classes of dynamic binary choice environments. We show that in these settings, focusing only on experiences that followed the same sequence of outcomes preceding the current task is more effective than focusing on the most recent experiences. Moreover, this process approximates the optimal strategy in many such settings. Next, we examine the psychological significance of these sequence-based rules. We show that these tractable rules reproduce well-known indications of sensitivity to sequences and predict a non-trivial wavy recency effect of rare events. Analysis of multiple published data-sets of learning experiments supports the wavy recency prediction. Specifically, the probability of risky choice after a rare favorable outcome from risk-taking is initially relatively high, but very soon after (within 2-3 trials) it falls to a minimum, then increases for about 12 trials, and then decreases again. Rare negative outcomes trigger a similar wavy reaction when the feedback is complete, but not under partial feedback. The difference among the effects of rare positive, rare negative, and non-rare outcomes and between full and partial feedback settings can be described as a reflection of an interaction of an effort to discover patterns with two other features of human learning: surprise-triggers-change and the hot stove effect. Similarity-based descriptive models are shown to capture well all these interacting phenomena, as well as the main behavioral phenomena documented in basic decisions from experience and probability learning tasks. We conclude with theoretical notes on similarity-based learning.