Individual differences in the selection and use of features for classifying visual stimuli

Area of Science:

• Cognitive Psychology
• Human-Computer Interaction
• Pattern Recognition

Background:

• Understanding human perception and classification of objects is crucial for fields like design and artificial intelligence.
• Previous research has explored shape perception, but the specific mechanisms of classifying complex, real-world objects like pots require further investigation.

Purpose of the Study:

• To investigate how human judges classify pot-like shapes based on varying features.
• To identify differences in feature weighting and classification strategies among human judges.
• To explore the role of quantitative versus all-or-none features in object classification.

Main Methods:

• Experiment 1: Seventy-two abstract pot-like shapes with varying quantitative ratios were classified by 20 human subjects and a computer program.
• Experiment 2: 256 shapes traced from real pots, featuring both quantitative and all-or-none characteristics, were classified by 15 judges.
• Analysis focused on inter-judge variability in feature importance and classification criteria.

Main Results:

• Significant differences were found in how judges weighted different features when classifying pot shapes.
• Judges could be categorized based on their distinct feature-weighting patterns.
• Variability extended to the use of all-or-none features, indicating diverse classification approaches.

Conclusions:

• Human judgment in classifying pot-like shapes is subjective and varies considerably between individuals.
• Cognitive mechanisms underlying feature selection and weighting differ among judges.
• These findings have implications for developing more nuanced computer vision systems and understanding human categorization processes.