Changing your mind about the data: Updating sampling assumptions in inductive inference.

Area of Science:

• Cognitive Science
• Psychology
• Machine Learning

Background:

• Inductive inferences rely on both sample content and sampling assumptions.
• Understanding how people revise beliefs about data generation is crucial for cognitive modeling.

Purpose of the Study:

• To investigate whether individuals can update their sampling assumptions when presented with new information.
• To determine if learners can reinterpret evidence based on revised beliefs about sample generation.

Main Methods:

• A property induction task was employed, where participants inferred property generalization from sample data.
• Sampling assumptions were manipulated using 'property sampling' and 'category sampling' frames.
• Experiments involved presenting initial frames, followed by sample data, and in some cases, retracting and replacing frames.

Main Results:

• Sampling frames (property vs. category) influenced property generalization patterns.
• Generalization was narrower under a property frame compared to a category frame.
• Participants successfully updated their sampling assumptions based on later-presented frames, adjusting their inferences accordingly.

Conclusions:

• Learners demonstrate the capacity to revise incorrect beliefs about data selection processes.
• Updated sampling assumptions lead to adjusted inductive inferences, highlighting cognitive flexibility.
• These findings have implications for understanding human learning and decision-making under uncertainty.