关键词: 信息处理偏见, 预测准确率, 信息源

Abstract: In a statistical sense, individuals might make perfectly correct judgments by using information sources that always offer the wrong predictions, by simply doing the opposite of what the source recommends. Yet despite the statistical feasibility of this option, consumers do not seem to routinely follow information sources that are consistently wrong. Drawing on existing decision making research, especially that related to the status quo effect and positive confirmation bias (PCB), we argue that consumers start with a desire to test the “rightness” instead of the “wrongness” of an information source, which makes them more sensitive to accurate predictions than to errors. Overlooking errors has little influence on information sources with a high forecasting accuracy but could lead to systematically poor predictions about information sources with low forecasting accuracy. However, as consumers gradually disconfirm the default hypothesis with repeated feedback information, they likely assign more weight to the disconfirming information, which lessens bias in their memory and recall of incorrect predictions. If they have access to a high accuracy information source though, learning cannot reduce their bias, because the correct forecasts only strengthen consumers’ beliefs about their default hypothesis. Therefore, we also propose that a longer learning period improves the forecast performance of consumers exposed to a low accuracy information source condition but not those exposed to a high accuracy information source. We also argue that individuals are more likely to develop PCB when following a more normative information source, thus informativeness bias might be heightened.
Three studies show that decision makers tend to follow information sources with high accuracy rates, even though an information source with low accuracy offers the same level of informativeness for decision making. This positive following bias can lead people to underestimate the value of information from low accuracy information sources. Learning can reduce this positive following bias phenomenon. Furthermore, the positive following bias is more likely to occur for a more normative information source. An alternative explanation for the positive following bias explanation might be variance in information could influence consumers’ assessments of its value, such that consumers infer that a low accuracy rate implies greater variance or less reliability. In turn, they may decide that incorrect predictions are abnormal and not informative. Therefore, as a comprehensive test of the informativeness bias explanation, we also rule out the influence of the information variance factor by introducing a zero variance s, that is, when a person consistently makes wrong predictions.
Our findings also contribute to judgment and choice research. First, to the best of our knowledge, no other research has addressed biased acquisitions and the use of high error rate information, despite recognition of the value of this type of information in prior theoretical analyses. We test for the existence of an informativeness bias in binary choices situations and examine potential moderators of this effect. Second, we extend PCB research by examining its role as an antecedent of an informativeness bias. Third, we find that learning, previous experience, and information source normativeness have not only direct effects for reducing informativeness bias but also help clarify research on the regulation of PCB.

Key words: positive following bias, prediction accuracy rate, information source