Confirmation bias: Prevalence and debiasing techniques
Master thesis
Published version
Permanent lenke
https://hdl.handle.net/10642/8012Utgivelsesdato
2019Metadata
Vis full innførselSamlinger
Sammendrag
‘Confirmation bias’ is a phenomenon that has been recognized by philosophers of science through centuries. The phenomenon is commonly defined as a tendency to favor confirmatory evidence in support of already stated hypotheses or response patterns summarized as “beliefs”, rather than disconfirmatory evidence. Such behavior involves missing potentially important truths. The phenomenon has been discussed in relation to numerous disciplines and other areas, such as law enforcement, politics, social media, psychology, and science. It has been suggested that the errors that exemplify the bias is to blame in cases from superstition to innocent convictions and even the event of Trump becoming the president of the United States. The research literature with respect to experiments demonstrating the phenomenon as well as experiments testing different debiasing techniques is reviewed, with a focus on the latter. In addition, a behavior-analytic approach is proposed with respect to an account of the behavior considered as showing the bias and to an explanatory model of such behavior. Moreover, two types of training are suggested and tested as debiasing techniques, namely training with explanatory feedback and training with correct/incorrect feedback. The extent to which skills generalized to novel exemplars was measured. Two participants disqualified for further participation as a result of demonstating to already have the skills of interest established in their repertoire, whereas considerable improvement was demonstrated in 13 out of the remaining 20 participants. Eleven participants also showed generalized skills
Beskrivelse
Master i læring i komplekse systemer