Knowledge updating in real-world estimation: Connecting hindsight bias and seeding effects

Area of Science:

• Cognitive Psychology
• Decision Science
• Behavioral Economics

Background:

• Hindsight bias describes the tendency for estimates to shift towards known outcomes.
• Traditionally viewed as a cognitive error, recent theories propose it stems from adaptive knowledge updating.
• Existing research on hindsight bias and seeding effects in estimation accuracy remains largely disconnected.

Purpose of the Study:

• To provide a framework connecting knowledge updating and hindsight bias in real-world estimation.
• To investigate the role of metric domain knowledge recalibration in hindsight bias.
• To experimentally examine the association between hindsight bias and improved estimation accuracy, including transfer learning.

Main Methods:

• Developed an experimental paradigm to study hindsight bias and estimation accuracy.
• Experiment 1: Induced hindsight bias to assess transfer learning.
• Experiment 2: Investigated if transfer learning can trigger hindsight bias.

Main Results:

• Demonstrated that inducing hindsight bias leads to transfer learning.
• Provided evidence that hindsight bias can be triggered by transfer learning.
• Established a direct link between knowledge updating processes and hindsight bias.

Conclusions:

• Hindsight bias is driven by adaptive learning processes, not solely cognitive error.
• Knowledge updating, specifically recalibration of domain knowledge, enhances estimation accuracy via transfer learning.
• Integrates research on hindsight bias and seeding effects, offering a novel perspective on estimation and learning.