Establish population-level independence of the observed human convention pattern

Determine whether the convention-gap structure and indirect-reference excess observed in the two human Hanabi corpora generalize independently at the population level, rather than reflecting shared sampling or player-population effects.

Background

The paper analyzes two human Hanabi-game corpora drawn from the same platform and convention community. Their category profiles and indirect-reference effects replicate closely, but the open data do not contain player identifiers, so the authors cannot assess player overlap or establish that the two samples represent independent draws from the broader population.

The unresolved issue is whether the replicated convention structure is genuinely population-level rather than a consequence of shared community membership, overlapping players, or other selection effects. The paper explicitly distinguishes replication across game records from population-level independence.