Human brains construct individualized global rankings from identical few-shot learning input.

Area of Science:

• Cognitive Neuroscience
• Computational Neuroscience
• Psychology

Background:

• Ranking is crucial for organizing information and managing cognitive load.
• Real-world ranking often relies on limited, few-shot learning of local relationships.
• The brain's mechanism for relational inference from sparse data is not well understood.

Purpose of the Study:

• To investigate how the human brain performs relational inference under few-shot learning conditions.
• To explore the role of inductive biases in shaping subjective global rankings.
• To examine neural representations associated with subjective ranking construction.

Main Methods:

• A preregistered behavioral study involving few-shot learning of local pairwise relationships.
• Magnetoencephalography (MEG) recordings to capture neural activity.
• Comparison of individual subjective rankings against ground-truth order and classical computational models.

Main Results:

• Individuals construct stable, self-consistent, yet idiosyncratic global rankings, diverging from objective order.
• These subjective rankings are not readily explained by classical transitive inference models.
• MEG data shows frontoparietal neural representations align with individual subjective rankings, not objective ones.

Conclusions:

• Human relational inference under sparse data is shaped by active, idiosyncratic inductive biases.
• The brain constructs and imposes global relational structures, demonstrating a generative cognitive process.
• Cognitive and neural mechanisms prioritize subjective relational inference when data is limited.