Learning With Selective Feedback : Effects on Performance and Coding of Unknown Outcomes

Sammanfattning: In experiential learning, one important source of information is the feedback that people receive on the outcomes of their decisions. Typically, however, feedback is systematically absent for many decisions and the actual experience of people may therefore be highly selective. It is thus surprising that research on the cognitive processes involved in human judgement and categorisation has not addressed the effects of learning with selective feedback. In this thesis, three studies are presented in which the effects of learning with systematically selective feedback were investigated. The hypothesis of constructivist coding was introduced in Study I, suggesting a cognitive mechanism for the processing of selective information. In the absence of external feedback people infer the most likely outcome, and then code this inference into memory as “internal feedback”. This internally generated feedback is stored and processed in the same manner as externally presented feedback and is used as a basis for beliefs about the characteristics of the environment. Results from Studies I, II, and III demonstrated support for constructivist coding under varied learning conditions. Study III investigated the effects on the beliefs of participants when they learn from feedback received only for positive decisions. Results indicated that the participants’ beliefs well reflected their actual, however selective, experience. When participants aimed to achieve good immediate outcomes, their experience became restrictive and biased, resulting in biased beliefs. In contrast, when the focus of participants was on long-term learning, their decisions produced a more representative experience and their beliefs came to reflect the actual structure of the environment. Biased beliefs were thus demonstrated to result from a sensitivity of participants to selectively available information. The present findings offer an understanding of the cognitive processes involved in learning from selectively absent feedback. The conclusions propose a sensitivity of participants to objectively experienced information in the forming of knowledge and beliefs. Further, when external information is absent, participants appear to rely on their knowledge and expectations to infer and code the most likely outcome, and use these stored inferences to form a coherent representation of the environment.