Measures of Bayesian Reasoning Performance on 'Normal' and 'Natural' Frequency Tasks

Area of Science:

• Cognitive psychology
• Decision-making research
• Human reasoning

Background:

• Bayesian reasoning studies often focus on error avoidance, like base-rate neglect.
• Less research has explored how different numerical formats (probability vs. frequency) affect absolute accuracy.

Purpose of the Study:

• To investigate if presenting Bayesian reasoning problems in frequency formats (normalized and natural) versus probability formats impacts absolute accuracy.
• To examine if age influences the effect of numerical format on reasoning accuracy.

Main Methods:

• Two studies were conducted comparing performance on Bayesian reasoning tasks presented in probability, normalized frequency, and natural frequency formats.
• Participants included older adults (≥60 years) and younger adults (18-29 years).

Main Results:

• Frequency formats, especially natural frequencies, led to reduced over-estimation compared to probability formats.
• An interaction between age and task format was observed, with older adults showing less sensitivity to format changes.
• No significant association was found between numerical format and the avoidance of specific reasoning errors like base-rate neglect.

Conclusions:

• Frequency-based representations, particularly natural frequencies, may enhance absolute accuracy in Bayesian reasoning by reducing over-estimation.
• The findings suggest that frequency formats do not necessarily engage distinct cognitive processes (System 2 reasoning) for error avoidance.
• Older adults' reasoning performance appears less affected by numerical format variations in these tasks.