Measuring Negotiation Traits with Collaborative Choice Data: A Two-Stage Latent Variable Model

2026-10-05

AI Summary

This study addresses the challenge of quantifying individual negotiation traits in collaborative decision-making, where multi-agent rewards are inherently confounded. To this end, we propose a two-stage latent variable probabilistic framework that integrates individual rewards and latent traits to generate team utilities and choice behaviors. The model accommodates covariates and enables individual-level psychometric assessment within structured tasks via Bayesian inference. Simulation studies confirm the model’s capacity to accurately recover trait parameters. Empirical applications further demonstrate that the proposed approach achieves significantly superior predictive accuracy compared to baseline methods, with estimated traits exhibiting strong correlations with external criteria. This work effectively fills a methodological gap by providing a principled approach to quantifying individual negotiation capabilities in collaborative settings.

Abstract

Consider a collaborative decision-making setting in which team members with different reward structures need to make a joint decision. Repeated collaborative decisions can reflect individual differences in team members' negotiation ability. However, extracting such information is challenging because each observed response arises from a collaborative process involving multiple participants and their reward structures. In this paper, we propose a two-stage probabilistic measurement model for structured collaborative choice tasks. In the first stage, known member-specific rewards and latent participant negotiation traits jointly determine team-option utilities, from which a latent team choice is generated. In the second stage, individual responses are modeled conditional on the latent team choice, with possible deviations that depend on participants' own rewards. Participant-level covariates can also be incorporated through a structural model. A simulation study shows satisfactory recovery of the participant traits and model parameters across different sample sizes and item sizes. An application to data from simulated collaborative negotiation tasks shows choice prediction accuracy well above the random-choice baseline and meaningful associations between the estimated traits and external criterion variables. The proposed framework provides a psychometric approach to measuring participant-level negotiation traits with collaborative choice data.

Problem

• collaborative decision-making
• negotiation traits
• latent variable model
• collaborative choice data

Innovation

• Two-stage latent variable model
• Collaborative choice data
• Negotiation traits
• Probabilistic measurement model
• Psychometrics