|Ph.D Student||Luria Adi|
|Subject||On the Effects of Immediate Feedback|
|Department||Department of Industrial Engineering and Management||Supervisor||Professor Ido Erev|
|Full Thesis text|
A large body of literature in psychology presents positive effects of immediate feedback on learning, and this dissertation presents some boundaries for these effects. The two chapters presented explore two common definitions of the feedback construct, namely 'knowledge of results' and 'reinforcement'. In the first chapter, feedback is utilized as knowledge of results in an abstract reaction-time task. This chapter explores the role of immediate “informative feedback” that provides information concerning the error made in the last response. A theoretical model is presented that can capture the sometimes mixed results regarding immediate feedback's effect on learning. The model suggests that the effect of immediate feedback is a joint outcome of two effects, one positive and one negative. The positive effect of immediate feedback is the close association between actions and their outcomes. The negative effect of immediate feedback is attributed to the tendency of immediate feedback to guide performance during acquisition, creating a dependency on feedback that harms performance once feedback is removed. Thus the net effect of immediate feedback is expected to be sensitive to manipulations that change the magnitude of the guidance effect and the importance of the acquisition rate. Four experimental studies presented in Chapter 1 evaluate this suggestion and offer some possible boundary effects.
In the second chapter, feedback is utilized as reinforcement in a computerized decision making paradigm. The chapter explores the value of delaying the immediate monetary feedback by replacing it with immediate symbolic reinforcement that provides probabilistic information. The chapter focuses on an environment in which the likelihood of monetary reinforcement is low. Previous decision making research suggests that in this setting the monetary reinforcement is likely to be underweighted; thus, we hypothesize that more frequent symbolic reinforcement can be effective. Study 1 evaluates the effectiveness of different frequencies of symbolic reinforcements. Results of Study 1 show that the rate of promoted behavior is a non-linear function of reinforcement frequency. Results of Study 2 suggest that there exist individual differences and varying preferences in assessing the reinforcing value of the symbolic reinforcement, meaning that no evidence for a general "uncertainty avoidance" pattern is evident. The chapters seek to clarify the psychological effects associated with immediate feedback, and suggest specific conditions under which providing immediate feedback may be problematic. Likewise, the studies seek to clarify conditions under which immediate feedback is expected to be an effective learning tool.