A teamwork response tree model for dyadic agreement-accuracy responses

Teamwork research increasingly relies on collaborative problem-solving (CPS) tasks to assess how individuals coordinate and integrate knowledge in joint work settings. However, most existing measurement approaches focus on final task accuracy and provide limited insight into the interaction processes that produce collaborative outcomes. Conventional scoring methods typically confound whether team members reach agreement with whether their responses are correct. This study introduces the Teamwork Response Tree (TRTree) model that decomposes dyadic collaborative responses into two linked latent processes: an agreement process and an accuracy process. The model is formulated within the item response tree framework and estimated using Bayesian methods. An empirical illustration using dyadic collaborative matrix reasoning tasks shows that the model yields interpretable estimates of individual ability, team-level ability, team-level agreement propensity, and item-level agreement difficulty. A Monte Carlo simulation study further evaluates parameter recovery under varying data conditions. Results indicate satisfactory parameter recovery and predictable effects of data informativeness on estimation accuracy. By separating convergence from correctness, the TRTree model provides a process-sensitive approach to measuring teamwork and extends item response tree modeling to collaborative problem-solving contexts.