Now showing 1 - 2 of 2
  • Thumbnail Image
    Some of the metrics are blocked by your 
    Item type:Publication,
    Two Tales of Uncertainty: Calibrating Overprecision With Self‐Regulated Entrepreneurial Learning
    Entrepreneurial success strongly depends on decision-making under uncertainty. Uncertainty is particularly pronounced whenentrepreneurs are confronted with novel situations characterized by an absence of information. While cognitive biases, such asoverprecision, can have a negative impact on decision-making processes by distorting information-processing, self-regulatedentrepreneurial learning (SREL) has been found to be beneficial when navigating the uncertain entrepreneurial environment,but its potential to calibrate overprecision remains unclear. This study addresses this gap by investigating whether an SREL in-tervention can calibrate overprecision, a highly detrimental form of overconfidence in entrepreneurship
    Type:
    Journal:
    Scopus© Citations 1
  • Thumbnail Image
    Some of the metrics are blocked by your 
    Item type:Publication,
    Cognitive Biases and Strategic Decision-Making: A Self-Regulation Perspective
    Entrepreneurs are prone to various cognitive biases which distort their learning and information processing. These biases vary considerably among entrepreneurs and impact information processing during strategic decision-making. While research has examined the outcomes of cognitive biases, there is a lack of knowledge on how cognitive biases impact entrepreneurial decision-making processes. To address this gap, we analyze how cognitive biases impact information processing during decision-making processes. Furthermore, we discuss how entrepreneurs can self-regulate the effects of overconfidence and attribution bias in more detail. We discern functions of cognitive biases across different entrepreneurial situations and suggest that entrepreneurs with high self-regulation are able to adapt to specific situations by monitoring and controlling their cognition and affect.
    Type: